So 127 (13) 12-2014.cdr

Tạp chí Khoa học và Công nghệ

SPECIAL ISSUE ON THE 49th ESTABLISHMENT OF COLLEGE OF TECHNOLOGY – TNU

(19/8/1965 – 19/8/20140

Content Page

The Quang Phan, Dung Thi Quoc Nguyen, Thao Thi Phuong Phan - Effects of Workpiece Hardness on Hard

Turned Surfaces of Alloy Steels 3

Lanh Van Nguyen, Loc Bao Dam - Direct MRAS based an Adaptive Control System for Twin Rotor MIMO

System 9

Nguyen Minh Y, Thang N.Pham and Toan H. Nguyen - A new approach for enery saving to household

customers based smartgrid technologies 15

Dinh Thi Gia, Tuan Manh Tran, Son Que Tran - Direct mras with safe constraints applied for two-wheeled

mobile robot 21

Phong Tien Le, Minh Duc Ngo - Research on designing an energy management system

for isolated pv source 29

Phong Tien Le, Huong Thi Mai Nguyen, Hung Tien Nguyen - Control of grid-connected solar power systems

with interleaved flyback converters 37

Nam Hoai Nguyen, Trinh Thi Minh Nguyen - A new training procedure for a class of recurrent neural networks 43

Cam Thi Hong Nguyen, Trang Van Nguyen, Pi Ngoc Vu - A new study on optimum calculation of partial transmission ratios of coupled planetary gear sets 47

Thuy Thi Thu Nguyen - Establishment of a database of emission factors for atmospheric pollutants from steel

rolling 51

Duy Tien Nguyen, Binh Hoang Lam, Son Hung Lam, Huy Phuong Nguyen - Dissolved Oxygen Control of the Activated Sludge Wastewater Treatment Process Using Hedge AlgebraicControl 57

Khuyen Thi Minh Pham, Yen Thi Mai Pham - Supply chain management for colleges/universities: solutions to

improve the efficiency of science and technology transfer 63

Thao Thi Phuong Phan, Thinh Duc Nguyen, Oanh Thi Lam Nguyen - Design and fabrication of robotic bluetooth cleaner 69

Thinh Duc Nguyen, Tuan Anh Vu, Du Van Nguyen- Caculation analysis and design for construction of solar

engine model 73

Ha Thi Thu Phan, Thao Thi Phuong Phan - Effect of annealing treatment on high strain rate behavior of

Graphene reinforced Polyurethane composites 77

Thuy Thi Hong Truong, Nga Thi Hong Do -Application of neural networks or diagnosis of hepatitis 81

Huy Ngoc Vu, Tuan Manh Tran, Huong Thi Mai Nguyen, Hung Tien Nguyen - Robust control of dc motors 87

Huy Ngoc Vu, Tuan Manh Tran, Huong Thi Mai Nguyen, Hung Tien Nguyen - Thyristor-based digital control of dc motors 93

Kien Ngoc Vu, Du Huy Dao, Cong Huu Nguyen - Research to improve the model order reduction algorithm 101

Viet Quoc Vu - Improving the efficiency of conventional drinking-water-treatment processes in the removal of

arsenic 107

Huyen Vu Xuan Dang, Hanh Vu Bich Dang, Amira Abdelrasoul, Huu Doan, Dan Phuoc Nguyen -

Assessment of treated latex wastewater reuse for perennial tree irrigation on ground water quality 111

Huyen Vu Xuan Dang, Huyen Thi Bich Trinh, Hanh Vu Bich Dang, Dan Phuoc Nguyen - Studying on toxicity of treated latex wastewater to plant perennial tree – a case study in Binh Duong, VietNam 117

Journal of Science and Technology 127(13)

2014

The Quang Phan et al Journal of SCIENCE and TECHNOLOGY 127(13): 3 - 8

3

EFFECTS OF WORKPIECE HARDNESS ON HARD TURNED SURFACES OF ALLOY STEELS

The Quang Phan, Dung Thi Quoc Nguyen* and Thao Thi Phuong Phan

University of Technology - TNU

ABSTRACT

Nowadays, hard turning is widely applied in Vietnam industry and it is usually the finished

operation so the quality of the machined surface plays a very important role to the use today and in

the future. This paper presents results of a research on hard turning of 9XC and X12M alloy steels

to explore the influence of workpiece’s hardness on machined surface roughness and topography

at selected cutting conditions. It is evident that the surface roughness was directly proportional to

the increase of the workpiece’s hardness from HRC = around 50 to higher than 60. Moreover,

lower hardness resulted in worse surface roughness. Even though when the cutting speed increased

by twice, the best surface roughness still achieved at the workpiece’s hardness of HRC= around

50. The cause is predicted to be involved with a change in chip/ rake face interactions depending

on workpiece’s hardness and tools wear.

Keywords: Hard turning, furface roughness, topography, workpiece, tool wear.

INTRODUCTION*

Precision machined components can be

manufactured by hard turned or ground

operations. Surface integrity is a qualitative

and quantitative description of both the

surface and subsurface component including

surface topography, surface and subsurface

hardness, microstructure and residual stresses,

etc. The work of Schwach and Gue [1] used a

stylus instrument to measured surface

roughness created by hard turn stated that

surface roughness decreased when feed rate

reduced. Decreasing feed rates makes the

surface residual stress more compressive and

its maximal one closer to the surface.

Moreover, tools wear increased surface

roughness except at moderate mode. Sharp

cutting tool is recommended for hard turn to

get better surface integrity. Chou [2] stated

that fine structure of the workpiece PM M50

steel resulted in lower wear rate by delay of

delamination wear and this effect is much

stronger in intermittent cutting.

Barbacki and co-workers [3] carried out

experiments to compare the microstructural

* Tel: 0915308818; Email: [email protected]

changes in the surface layer of hardened steel by

hard turning and grinding found that both

operations offered high surface quality of the

machined components. According to them,

favorable surface integrity can be achieved both

technologies and properly way to apply. Several

parameters such as thickness of white layer, its

hardness and stress level can be determined as a

function of cutting parameters and tools wear.

Kishawy and Elbestawi [4] studied effects of

process parameters on material side flow

during hard turning showed the formation of

material side flow based on two possible

mechanisms. First, the workpiece material

was squeezed between flank face and the

machined surface and it is clear when chip

thickness is less than minimal chip thickness.

Second, under high pressure and temperature,

the plastically deformed material was pressed

aside. The trailing edge notch was caused by

the chip edge serration. They also found that

feed rates, tools wear, tool nose radius and

edge preparation all have effects on material

side flow and of course on surface

topography. The formation of white layer on

the machined surface of hard turning was

studied by Chou and Evans [5], they stated


4

that the surface layer consists of two layer the

white outmost and dark layer just below. The

formation of white layer involves dominantly

with a rapid heating – cooling process. Plastic

deformation also helps grain refinement and

phase transformation to facilitate its

formation.

The study in this paper concentrated on the

effects of workpiece’s hardness on the surface

integrity particular on surface roughness and

its topography in the relation with certain

cutting conditions and tools wear.

EXPERIMENTAL PROCEDURE

Tool and Machine tool

The tools used in the study were PCBN equal

triangle inserts made in Korea. Machine tool

is a turning center CNC-HTC2050 made in

China. The tool was set up on tool handle and

then on the machine with: rake angle = - 6;

flank angle = 6; clear angle: 1 = 2 = 30.

Workpiece

Two types of workpieces were used namely

X12M and 9XC hardened steels (Russian

standards). Their chemical compositions were

analyzed by spectrographic method shown in

table 1 and 2. The hardness of the two

workpieces was divided into three categories:

HRC=4750; HRC=5457 and HRC=6063.

The microstructures of the two types of

steels were analyzed on optical microscopy

corresponding to the three categories of

hardness shown in Figure 1. When the

hardness of X12M steel increased from HRC

4750 to 5457 and 6063, the carbides

were observed to be elongated in shape,

concentrated in lines and increased from 3-5

µm to 10-25 m with high density.

However, the carbides in 9XC steels kept

quite stable with small size of approximately

1 µm when the hardness increased from

HRC 47 to HRC 63.

Table 1. Chemical composition of X12M steel

Element C Si P Mn Ni Cr Mo

Percentage % 1,4916 0,3589 0,0112 0,2404 0,2125 11,393 0,3803

Element Cu Ti Al Fe V

Percentage % 0,3383 0,0063 0,0249 85,396 0,1799

Table 2. Chemical composition of 9XC steel

Element C Si P Mn Ni Cr Mo

Percentage % 0,823 1,2351 0,0241 0,5862 0,0332 1,113 0,0192

Element Cu Ti Al Fe V

Percentage % 0,2876 0,1768 0,0299 0,0011 95,447 0,1499

Figure 1. The microstructure of X12M (a, b, c) and 9XC (a’, b’, c’) steels with the hardness approximately

HRC=4750; HRC=5457 and HRC=6063, respectively

a) b) c)

a’) b’) c’)


5

Cutting conditions

The cutting conditions were selected as

follows:

Cutting speed: v1 = 110 m/p; Feed rate: s1 =

0.12 mm/rev; un-depth of cut: t1 = 0.15 mm.

Cutting speed: v2 = 220 m/p; Feed rate: s2 =

0.12 mm/rev; un-depth of cut: t2 = 0.15 mm.

RESULTS AND DISCUSSION

Surface integrity

Surface roughness

When the first cutting condition was applied

the surface roughness measured by stylus

surface roughness divide, Mitutoyo SI-201

showed that the surface roughness was better

for 9XC steel compared with X12M in the

range Ra = 0.55 – 1.06 µm and Ra = 0.75 –

1.37 µm, respectively. The trends of surface

roughness of the two types of steels are

shown in Figure 2. It is clear that the higher

hardness of the steel was, the higher surface

roughness was. The surface roughness was

the lowest at the hardness of the workpiece of

HRC= 4750 with the value of approximately

Ra =0.55 µm. This result kept the same when

cutting speed increased by double value (the

second cutting condition). It is very

interesting to note that when lower workpiece

hardness was applied (HRC=40-43) for

testing both 9XC and 12XM steels, the

surface roughness was much higher than at

the hardness of HRC= 4750 with the value

around Ra=0.75 µm and 0.91 µm,

respectively. An effect of a change type of

chip formation at the workpiece’s hardness of

HRC=4750 might be the major factor.

Moreover, the longer cutting time was, the

higher surface roughness was, especially

when the cutting time increased by three

times, the surface roughness could increase

nearly twice. This indicated that tools wear

has strong effect on increasing surface

roughness.

The surface topography was taken on

Scanning Electron Microscopy (SEM) shown

in Figure 3 with different workpiece hardness

in the range HRC=4345; HRC=4750 and

HRC=6063. It is very clear that the side

effects are more serious at figure 3(a,c) and

much less effect in Figure 3(b) leading to the

best surface finish in this case. In Figure 3(a),

the type of plastic deformation in chip

formation is predominant and in Figure 3(c),

the type of cleavage in chip formation is

clearly observed. The evidence in Figure 3

supports for the ideas of a change of chip

formation at the workpiece’s hardness of

around the value of HRC=50. The ploughing

effect to smear work material on the

machined surface is also evident in this

figure.

Figure 2. Graphs showing increases of surface roughness of 9XC and X12M hardened steels depending on the

workpiece’s hardness; cutting speed: v1 = 110 m/p; feed rate: s1 = 0.12 mm/rev; un-depth of cut: t1 = 0.15 mm

m


6

Figure 3. SEM micrographs showing the surface topography after hard turning of X12M steel with

different hardness of workpiece: HRC=4345; HRC=4750 and HRC=6063; Cutting speed: v1 = 110

m/p; Feed rate: s1 = 0.12 mm/rev; un-depth of cut: t1 = 0.15 mm

The micro-hardness measurements on cross

section of the workpiece from the depth of 15

µm to 300 µm showed evidence the effects of

smearing on the machined surface resulting in

an increase in surface hardness at a very

narrow layer with the depth less than 15 µm.

It is reasonable because the depth of cut here

is quite small t = 0.15 mm at the level of

precision cutting and consistent with other

authors’ results.

Frictional Interactions between chip and rake face

It is evident in Figure 4(a) that at low

workpiece’s hardness (HRC=4345), the

length of contact is the longest (l = 300 µm)

and mainly covered by the work material.

However, the length of contact is reduced by

a half (l = 150 µm) shown in Figure 4(b)

when workpiece with the hardness of

HRC=5054 were machined. The rake face is

nearly free of material transfer. Moreover,

when the hardness of the workpiece was

HRC=6063, the length of contact increased

gain as shown in Figure 4(c) with l = 280 µm.

The main different compared with Figure 4(a)

is that material transfer is much less and

concentrated on the rear rake face. From

evidence in Figure 4, it is clear that there is a

change in frictional interactions between chip

and tool from mainly plastic type to cleavage

one in chip formation when the hardness of the

workpiece varied from around HRC=45 to 60.

Figure 4. SEM micrographs showing the rake face of PCBN inserts after hard turning of X12M steel with

different hardness of workpiece: HRC=4345; HRC=5054 and HRC=6063; Cutting speed: v1 = 110

m/p; Feed rate: s1 = 0.12 mm/rev; un-depth of cut: t1 = 0.15 mm

a) b) c)


7

Discussion

From the results mentioned above in the study,

the best surface roughness (Ra = approximately

0.55 µm) was achieved when X12 and 9XC

steel with the hardness of HRC = around 50

was machined by the first and second cutting

conditions. With the hardness HRC = around

45 and higher than 55, the surface hardness

was much worse. The fact can be explained by

the change in chip formation from plastic type

toward cleavage type similar to machining

brittle materials as shown in Figure 4. This

also involves with the type of frictional

chip/rake face interactions. Short chip/rake

contact and free of material transfer results in

low surface roughness and better surface

topography. Long chip/rake face length of

contact and more material transfer in both near

the cutting edge and at the region where chip

breaks from contact with the rake face cause

the higher surface roughness and worse surface

topography. This is completely consistent with

the ideas that the length of chip/rake face

contact is directly proportional to the value of

cutting force and surface roughness as a result

of the level of adhesion between chip and tool.

The hardness of the workpiece could change

the frictional contact on the rake face. When

the hardness reached HRC=6063, the first

crater with short length of contact formed

near the cutting edge and then the second

crater appeared at the rear of the first crater.

The harder of the chip shortened the length of

chip/tool contact on the rake face and after a

while when the crater developed enough it

formed the second one due to the depth of the

first crater changed the frictional contact on

the rake face.

CONCLUSION

From this study, conclusions can be derived

as follows:

The surface integrity estimated by surface

roughness and surface topography is consider

to the best for both type of workpiece

materials at the hardness HRC= 4750. The

surface topography shows that at low

hardness of HRC = 4750 chip formation

mainly in plastic type and at high hardness of

HRC = 55 and above the chip formation

changed toward cleavage similar to brittle

materials in cutting.

The frictional chip/tool interactions are also

changed depending on the workpiece’s

hardness. The lower hardness the longer

chip/tool contact is with full of material

transfer on the contact area. However, when

the hardness of the workpiece is higher than

HRC = 55, the contact length is shortened

with free material transfer and after a duration

of cutting, the second crater appears at the

rear of the first crater with not much material

transfer.

REFERENCES

1. D.W. Schwach and Y.B. Guo.; “Feasibility of

producing optimal surface integrity by process

design in hard turning”, Materials Science and

Engineering, A 395 (2005), pp. 116-123.

2. Y.K. Chou., “Hard turning of M50 steel with

different microstructure in continuous and

intermittent cutting”, Wear 255 (2003), pp. 1388-

1394.

3. A. Barbacki, M. Kawalec, A. Hamrol.,

“Turning and Grinding as a source of of

microstructural changes in the surface layer of

hardened steel, Journal of Materials Processing

Technology, 133 (2003), pp. 21-25.

4. H.A. Kishawy and M.A. Elbestawi., “Effects

of process parameters on materials side flow

during hard turning”, International Journal of

Machine Tools & Manufacture, 39 (1999), pp.

1017-1030.

5. Y.K. Chou and C.J. Evans., “White layers and

thermal modeling of hard turned surfaces”,

International Journal of Machine Tools &

Manufacture, 39 (1999), pp. 1863 -1881.

6. N.T.Q. Dung., “A study of hard turning

process with the use of PCBN inserts”, PhD

Dissertation, Thai Nguyen University of

Technology, 2012.


8

TÓM TẮT

ẢNH HƯỞNG CỦA ĐỘ CỨNG PHÔI

TRONG QUÁ TRÌNH TIỆN CỨNG THÉP HỢP KIM

Phan Quang Thế, Nguyễn Thị Quốc Dung*, Phan Thị Phương Thảo

Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên

Hiện nay, công nghệ tiện cứng đã được ứng dụng rộng rãi trong công nghiệp ở Việt Nam.

Tiện cứng thường là quá trình gia công lần cuối nên chất lượng bề mặt gia công đóng vai trò rất

quan trọng đối với việc sử dụng công nghệ tiện cứng trong hiện tại và tương lai. Bài báo này trình

bày kết quả một nghiên cứu về quá trình tiện cứng thép hợp kim 9XC và X12M nhằm xác định

ảnh hưởng của độ cứng phôi đến hình học và nhám bề mặt gia công trong điều kiện công nghệ xác

định. Kết quả cho thấy trong dải độ cứng từ 50 đến 60HRC nhám bề mặt tỉ lệ thuận với độ cứng

phôi. Tuy nhiện ở độ cứng thấp hơn chất lượng bề mặt giảm và nhám bề mặt tăng. Nhám bề mặt

đạt giá trị tốt ở độ cứng xấp xỉ 50HRC ngay cả khi tốc độ cắt tăng gấp đôi. Hiện tượng này được

cho là có liên quan đến việc thay đổi tương tác tiếp xúc giữa phôi và mặt trước của dao phụ thuộc

vào độ cứng phôi và mòn dụng cụ.

Từ khóa: Tiện cứng, nhám bề mặt, hình học, phôi, mòn dụng cụ.


Lanh Van Nguyen et al Journal of SCIENCE and TECHNOLOGY 127(13): 9 - 14

9

DIRECT MRAS BASED AN ADAPTIVE CONTROL SYSTEM FOR TWIN ROTOR MIMO SYSTEM

Lanh Van Nguyen*, Loc Bao Dam

University of Technology – TNU

ABSTRACT

In this paper, a Model Reference Adaptive Systems (MRAS) based an Adaptive System is

proposed to a Twin Rotor MIMO System (TRMS). The TRMS is an open-loop unstable, non-

linear and multi output system. The main task of this design is to keep the balance and to track a

given trajectory. There are two separate adaptive controllers designed for controlling two angles.

By applying Lyapunov stability theory the adaptive law that is derived in this study is quite simple

in its form, robust and converges quickly. Experimental results show that the proposed adaptive

PID controllers have better performance compared to the conventional PID controllers in the sense

of robustness against internal and/or external disturbances.

Index Terms – Model Reference Adaptive Systems (MRAS), Twin Rotor MIMO System (TRMS).

Keywords: Model Reference Adaptive Systems (MRAS), Twin Rotor MIMO System (TRMS).

INTRODUCTION*

The TRMS which isa model of the simpli fied

heli copter. Its position and velo city are

controlled by changing the speed of pitch and

yaw rotors. The TRMS system has high non

line arity, uncer tainty, especially coupling

between input sandout puts. This would be

avery complicated problem if we want to

control the TRMS moving quickly and

accurately to the desired location or a given

trajectory. The motion control system can

bequite complex because many different

factors must be conside redin the design. It's

hard to figure out the design methods that

consider all the factors such as: reducethe

effects of noise, object variable parameters,

avoid the influence of the coupling. There is

nosing lesolution to this problem.

There have been many research papers in

order to control the system. How ever the

classic controller will notachieve the desire

dresults. There fore, advanced controller was

introduced.

In this study, design of MRAS-based adaptive

control systems is developed for the TRMS

which acts on the errors to reject system


disturbances, and to cope with system

parameter changes. In the model reference

adaptive systems the desired closed loop

response is specified through a stable

reference model. The control system attempts

to make the process output similar to the

reference model output.

Fig 1: Experimental setup

Figure 1. Experimental setup

The proposed controller is expected to

improve the tracking performance and

increase the robustness under the effects of

disturbances and parameter changes. Two

separate adaptive controllers are designed

based on the Lyapunov’s stability theory for

controlling two given trajectory.

This paper is organized as follows. Design of

MRAS based an adaptive controller is


10

introduced in Section II. In Section III, the

dynamics of the twin rotor MIMO system is

shown. The design of the proposed controller

is introduced in Section IV. The experimental

results are also presented in section V. At the

end of this paper, summary of the paper is

given.

DESIGN OF DIRECT MRAS

Figure 2. Adaptive system designed with Liapunov

The structure depicted in Fig 1 can be used as

an adaptive PID controlled system. A second-

order process is controlled with the aid of a

PID-controller. Variations in the process

parameters bp, ap and Kw can be compensated

for by variations in parameters of this

controller Kp, Kd and Ki. We are going to find

the form of the adjustment laws for Kp, Kd and

Ki. The following steps are thus necessary to

design an adaptive controller with the method

of Lyapunov:

1. Determine the differential equation for :

= , (1) where and are

states of the reference model and process,

respectively.

2. Choose a Lyapunov function :

= , (2) in which a positive

definite symmetrical matrix, a diagonal

matrix with in principle arbitrary coefficients

0, and is the parameter error vector.

3. Determine the condition under which is

definite negative.

4. Solve from , (3)

where is the process matrix and is a

positive definite symmetrical matrix. This

yields, the form of the adjustment laws [2]:

(4)

In Equation 4 , and are called the

adaptive gains, and , , , , and are

defined in Fig 2, , are elements of the

matrix.

TWIN ROTOR MIMO SYSTEM

In order to design a controller for the TRMS,

a dynamical model is first required [3].

DESIGN OF CONTROL SYSTEM

PID Control System with Fixed Parameters


11

The PID control algorithm is mostly used in

the industrial applications since it is simple

and easy to implement when the system

dynamics is not available. For the TRMS

control variables are a pitch angle and yaw

angle such that two separate controllers

are required. In this study, the PID controller

is used for the given trajectories control.

There are many methods of choosing suitable

values of the three gains to achieve the

satisfied system performance. In this study,

the Ziegler – Nichols approach is used to

design PID controller to achieve a desired

system performance.

+

+

Twin Rotor

MIMO

System

-

PID1

PID2

-

Fig 3: PID controller structure

Figure 3. PID controller structure

Adaptive PID Control System

For purposes of comparison, the process is

repeated using an adaptive control structure.

The pitch angle and yaw angle of the TRMS

are controlled separately by two adaptive

controllers by replacing two corresponding

linear controllers indicated in Fig. 4.

Reference Model

Reference model is described by the transfer

function:

(6)

The parameters of the reference model are

chosen such that the higher order dynamics of

the system will not be excited. This leads to

the choice of

and , such that:

(7)


12

Figure 4. Adaptive PID control structure

State Variable Filter

As mentioned in Section II, the derivative of

the error can be created using a state variable

filter. The parameters of this state variable

filter are chosen in such a way that the

parameters of the reference model can vary

without the need to change the parameters of

the state variable filter every time. The

parameters are chosen as:

, and , then

(8)

Adaptive Controllers based on MRAS

Follow Ep. (4) the complete adaptive laws in

integral form for the pitch angle controller are

(9)

For the yaw controller

(10)

In the form of the adjustment laws ,

, and are elements of the

and matrices, obtained from the solution of

the Lyapunov equations indicated in Eq. (11):

;

(11)

- + +

- +

Reference Model 2

- +

+ -

+ + +

- Lyapunov

SVF

SVF

+ -

- +

Reference Model 1

- +

+ + -

Twin Rotor MIMO System

+ + +

-

-

-

+ +

Lyapunov

+ +


13

where and are positive definite

matrices and and are:

(12)

With , ;

and are adaptive

gains.

EXPERMENTAL TESTS

From the experimental results with two sets

of PID controller and adaptive PID controller

in Fig 5 we find that, the system using

adaptive PID controller has result in sticking

and eliminates noise better than that useing

the classical PID one.

Figure 5. Responses of the PID control system (left hand side)and adaptive PID control (right hand side)

system with disturbance

Fig 6. Adaptive PID parameter


14

CONCLUSION

In this paper, the conventional PID controller

and the adaptive PID controllers are

successfully designed to TRMS under

disturbances. The simple adaptive control

schemes based on Model Reference Adaptive

Systems (MRAS) algorithm are developed for

the asymptotic output tracking problem with

changing system parameters and disturbances

under guaranteeing stability. Experiments

have been carried out to investigate the effect

of changing the external disturbance forces on

the system. Based on the experimental results

and the analysis, a conclusion has been made

that both conventional and adaptive

controllers are capable of controlling the

given trajectory of the non-linear system.

However, the adaptive PID controller has

better performance compared to the

conventional PID controller in the sense of

robustness against disturbances.

REFERENCE

1. Van Amerongen, J., Intelligent Control (part

1)-MRAS, Lecture notes, University of Twente,

The Netherlands, March 2004.

2. Nguyen Duy Cuong, Nguyen Van Lanh, Dang

Van Huyen, “Design of MRAS-based Adaptive

Control Systems”, The IEEE 2013 International

Conference on Control, Automation and

Information Sciences (ICCAIS), pp. 79 - 84, 2013.

3. Twin Roto MIMO System Control

Experiments 33-949S Feedback Instruments Ltd,

East susex, U.K., 2006.

TÓM TẮT

HỆ THỐNG THÍCH NGHI MÔ HÌNH MẪU TRỰC TIẾP DỰATRÊN HỆ THỐNG ĐIỀU KHIỂN THÍCH NGHI CHO HỆ THỐNG TWIN ROTOR MOMO

Nguyễn Văn Lanh*, Đàm Bảo Lộc

Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên Bài báo này, đề xuất một hệ thống thích nghi theo mô hình mẫu (MRAS) đã được áp dụng vào hệ

thống Twin Rotor MIMO (TRMS). TRMS là một hệ thống hở không ổn định, phi tuyến có nhiều

đầu vào/ra. Mục đích chính của thiết kế này nhằm giữ cho hệ thống cân bằng và chuyển động bám

theo một quỹ đạo cho trước. Để thực hiện thiết kế cần thực hiện qua các bước sau: Bước 1, xây

dựng hệ phương trình chuyển động của đối tượng dựa theo phương trình Lagrange. Bước 2, thực

hiện tuyến tính hóa các phương trình. Bước 3, thiết kế hai bộ điều khiển thích nghi độc lập để điều

khiển hai góc đầu ra. Luật điều khiển thích nghi áp dụng lý thuyết ổn định Lyapunov có dạng đơn

giản, bền vững và hội tụ nhanh. Các kết quả mô phỏng và thực nghiệm chỉ ra rằng các bộ điều

khiển PID thích nghi có chất lượng tốt hơn khi so sánh với các bộ điều khiển PID thông thường

dưới tác động của các nhiễu bên trong và/ hoặc nhiễu ngoài.

Từ khóa: Hệ thống thích nghi theo mô hình mẫu (MRAS), Hệ nhiều đầu vào nhiều

đầu ra Twin Rotor (TRMS).


Minh Y Nguyen et al Journal of SCIENCE and TECHNOLOGY 127(13): 15 - 20

15

A NEW APPROACH FOR ENERY SAVING TO HOUSEHOLD CUSTOMERS BASED SMARTGRID TECHNOLOGIES

Minh Y Nguyen*, Thang N. Pham and Toan H. Nguyen


ABSTRACT

This paper addresses the energy efficiency problem of household customers by observing and

responding accordingly to the condition of the upstream grid; the key condition is the market price

which is passed to the end-use customers though a new market entity, namely load aggregators. A

framework based on Smargrid technologies, e.g., Advanced Metering Infrastructure (AMI) for

monitoring home energy consumptions is proposed. The problem is to schedule and control the

home electrical appliances in response to the market price to minimize the energy cost over a day.

The problem is formulated using Dynamic Programming (DP) and solved by DP backward

algorithm. Using stochastic optimization techniques, the proposed framework is capable of

addressing the uncertainties related to the appliance performance: outside temperature and/or

users’ habits, etc.

Keywords: Demand response, Home energy efficiency, Heat ventilation and air conditioning,

Dynamic programming, Smartgrid.

INTRODUCTION*

This paper discusses a new approach to

energy efficiency in the residential sector by

watching the household consumption from

the system perspective: it is more economical

and efficient not only for household

customers but the system-wide if the

appliances and lighting are turned on in low

price times and off in the high time. This can

be referred to as Demand Response (DR)

program and/or Home Energy Management

System (HEMS). Herein, we propose a DR

framework for a household that consists of

two functions: (1) Off-line scheduling

according to the prediction and (2) On-line

control based on both the previous scheduling

and real-time load measurements. The

framework is based on advanced

communication and automation technologies

applied to the power grid, i.e., Smartgrid with

the key component is Advanced Metering

Infrastructure (AMI). The problem is finding

out the optimal consumption each time slot


(stage) of the day to minimize the overall

energy cost, subject to the constraint of the

physical system and the users’ preference of

comfort.

THE PROPOSED DEMAND RESPONSE

FRAMEWORK

The proposed DR framework for household

customers is sketched in Fig. 1. As

aforementioned, under market environments

electric customers are offered choices to pay

their usage corresponding to the condition of

the wholesale market, i.e., real-time price.

The matter of fact is that it is not suitable for

human to analyze and respond to the frequent

change over time of the real-time price (e.g.,

every 5 min.). Therefore, advanced

communication and automations, also known

as Smartgrid technologies are essentially

needed here. The proposed scheme consists of

two different functions: (1) Off-line

scheduling according to the anticipated price

and load models and (2) On-line control

combining both the scheduling ahead and the

real-time measurements.


16

Figure 1. The DR framework based on AMI for household customers

Day-ahead Scheduling

The problem of day-ahead scheduling is to

find out the so-called “control policy” that

minimizes the expected energy cost over a

day with respect to the uncertainties of the

forecasting. The solution is subject the

constraints associated the physical system

capacity and/or the users’ preference of

comfort, etc. It is worth noting that the

decision is made in accordance with the time

basis of the electricity market, which is 5 min.

in this paper. The problem can be formulated

as follows [7].

1

0 0,1... 1 0,1... 1

min , ,k k

N

N N k k k kku wk N k N

E g x g x u w

(1)

Subject to

1 , , , 0,1... 1k k k k kx f x u w k N (2)

min max , 0,1... 1ku u u k N (3)

min max , 1, 2...kx x x k N (4)

1,2...

, , 0, 1,2...i k k k k Nh x u w i n

(5)

where xk the state variable at the beginning of

stage k; uk the control variable during stage k;

wk the uncertainty during stage k; N is the

number of stages over the scheduling period;

gN(xN) is the terminal cost, i.e., the cost

associated with the final state; gk(xk,uk,wk) is

the cost in stage k; fk(xk,uk,wk) is the state

transition function; umin, umax are the capacity

limits; xmin, xmax are the physical limits of the

system; and hi(x1,x2…xN) refers to the

customers’ preference, e.g., human would feel

comfortable if the temperature is kept within

22–260C with HVAC; batteries must be full

of charge by 7:00 AM (i.e., the time to go

working) with EVs.

In this formulation, equation (1) is the

objective function, i.e., minimizing the energy

costs over a day subject to the uncertainties.

Equation (2) shows the modeling of loads

which represents the dynamics (state

transition) of the load performance. Equations

(3) and (4) are the physical constraint of the

loads. Finally, equation (5) represents the

conditions to be comfortable setup by

customers and n is the number of functions

needed.

Figure 2. The proposed DR framework and its

variables defined in stage k


17

The control policy resulted from the day-

ahead scheduling is a set of functions of the

system state, denotes μk(xk), k = 0, 1…N-1,

that will point out the optimal control of the

system provided the measurement of the

current state (i.e., the scenario is cleared).

Real-time Controller

The fact is that the scenario probably differs

from the anticipation due to many uncertain

factors, e.g., weather, temperature and users’

demands, etc. Therefore, the real-time control

should not only follow the previous schedule

but also depend on the real-time measurement

of the system. With the control policy

determined ahead of time, the decision in

real-time operation can be made as simply as:

*

k k ku x (6)

where xk is measures of the state variable.

The block diagram of the proposed DR and its

variables defined in stage k are expressed in

Fig. 2.

ILLUSTRATIVE EXAMPLE

This section provides an illustrative example

where the proposed framework is tested in the

DR problem of HVAC loads in a real-time

electricity market. The idea of HVAC is

taking advantages of the slow dynamics of the

heating/cooling process compared to the

changing rate of the price signal (i.e., 5 min.)

to manage the HVAC operation with the

target of minimizing the total energy cost in a

day while maintain comfort levels to the

users. The framework specified for HVAC is

as follows.

1

0 0,1... 10,1... 1

min ac

kk

N ac

k kkTqk Nk N

E q

(7)

Subject to

1

1

or

1 , 0,1... 1

ac

k k k k k

ac

k k k k

t t q T t

t t q T k N

(8)

min max , 0,1... 1ac ac ac

kq q q k N (9)

min max , 1,2...kt t t k N (10)

whereac

kq is the energy consumption of

HVAC during state k, [kWh]; tk is the indoor

temperature at the beginning of stage k, [0C];

Tk is the outside average temperature during

stage k, [0C]; N is the number of stages; α is

the equivalent thermal resistance of HVAC,

[0C/kWh]; β is the coefficient of the heat

transfer between the indoor and outdoor

space, [p.u.]; min max,ac acq q are the capacity limits

of HVAC, [kWh]; and tmin, tmax are the lower

and upper temperature of human comforts,

[0C], e.g. human feels comfortable with the

temperature between 22–260C. It is worth

noting that with time basis of electricity

market is 5 min., resulting in the number of

stages is 24×12 = 288 in a day. The control policy constructed through the above scheduling problem combined with the real-time measurements (of the actual indoor temperature) will be used to determine the optimal decision as follows.

*

k k kq t (11)

The proposed HVAC scheme will be compared with the traditional scheme that HVAC is controlled by a thermostat. With the upper/lower set-points, HVAC will be switched on/off when the indoor temperature reaches the lower or upper bound of the customers’ preference, respectively. This can be described mathematically in the following.

max min

1 min max

max

if

if

0 if

ac

k

ac ac

k k k

k

q t t

q q t t t

t t

(12)

The price data used in this study is obtained by modifying the hourly electricity price of PJM market from March 24th to 30th 2014 (Monday to Sunday of a whole week) [8]. It is noted that the hourly price is determined through day-ahead bidding in electricity markets. The real-time imbalance caused by load deviations from the anticipation will be handled by calling upon the up/down regulation services; this action results in the real-time price differed from the hourly price [9]. The temperature in this study is referred from National Climate Data Center at New York, USA in the same period as the PJM price [10]. The price and temperature data are displayed in Fig. 3.


18

0 24 48 72 96 120 144 1680

0.1

0.2

0.3

0.4

Time (hour)

Pri

ce (

$/k

Wh

)

0 24 48 72 96 120 144 168-20

-10

0

10

Time (hour)

Tem

p.

(0C

)

Figure 3. The PJM hourly price and temperature

in New York, USA from March 24th to 30th 2014

Fig. 4 illustrates the control policy at some

stages (stage No. 5, 6, and 100) on March 24th

2014: the optimal decision, *

kq [kWh] as a

function of the state variable, tk [0C]. Two key

drivers of the control policy at each stage are:

(1) the price signal and (2) the temperature of

the following stages; thus, the HVAC tends to

run a little at stage No. 6 (green, dashed line)

since the price is quite high and expected to

drop soon; in contrast, at stage No. 100 (red,

broken line) HVAC is operated intensively to

drive up the indoor temperature, avoiding to

run with the high price in the coming stages.

It is noted that the maximum capacity of

0.5kWh per stage is equivalent to a 6kW

power drawn from the grid.

22 23 24 25 26

0

0.2

0.4

0.6

tk (0C)

q* k (

kW

h)

At stage No. 5

At stage No. 100

At stage No. 6qmax

= 0.5kWh

Figure 4. The control policy on March 24th 2014

obtained from the scheduling problem

Fig. 5 shows the indoor temperature of

HVAC controlled by the proposed and

traditional scheme on March 24th 2014. As

aforementioned, the traditional scheme is

based on a thermostat to drive the indoor

temperature from the lower to upper bound

(red, broken line), repetitively. On the other

hand, the proposed scheme (blue, solid line)

considers both the trend of electricity price

and temperature in the whole period (a day);

and the control policy is constructed by

minimizing the total energy cost subject to the

uncertainties of the prediction (e.g., the

outside temperature). Thus, the HVAC will

run with different operating levels throughout

the day, driving the indoor temperature within

the comfortable range (22–260C). The

economic performance of the proposed

scheme in comparison with the tradition

scheme is expressed in Fig. 6.

0 2 4 6 8 10 12 14 16 18 20 22 2421

22

23

24

25

26

27

Time (hour)

t k (

0C

)

The proposed scheme

The traditional scheme

Figure 5. The indoor temperature with the proposed

and tradition operation scheme on March 24th 2014

3/24 3/25 3/26 3/27 3/28 3/29 3/300

1

2

3

4

5

6

Day(month/day)

Co

st (

$)

The proposed scheme

The traditional scheme

Figure 6. The cost per simulated day (from March

24 to March 30 2014)


19

Fig. 6 shows the energy cost in each day of

the simulated period: from March 24th to 30th

2014. Generally, the cost is quite high from

March 24th to 27th due to two reasons: first,

the weather is cold with the temperature is

usually lower than –10 Degree Celsius and

secondly, the electricity price is relative high

in weekdays (from Monday to Thursday),

particularly the critical peak price occurs on

Thursday March 27th. In contrast, the cost

from March 28th to 30th is much slower

because the temperature rises significantly

(0–10 Degree Celsius) and the electricity

price also decreases somehow in the

weekend.

As the simulation results, it can be recognized

that significant saving can be obtained by the

proposed DR scheme on HVAC loads

compared to the traditional operation.

Particularly with the critical peak price on

Thursday, the proposed scheme can manage

the energy cost to be not increased that much

and obtain the highest saving throughout the

simulated week. In overall, the energy cost

with the proposed scheme is 18.67$ while that

with the traditional scheme is 21.85$; the

saving in this case is about 14.55%. The

HVAC modeling parameters and the

customer preference is provided in Table 1.

As the simulation results, it can be recognized

that significant saving can be obtained by the

proposed DR scheme on HVAC loads

compared to the traditional operation.

Particularly with the critical peak price on

Thursday, the proposed scheme can manage

the energy cost to be not increased that much

and obtain the highest saving throughout the

simulated week. In overall, the energy cost

with the proposed scheme is 18.67$ while that

with the traditional scheme is 21.85$; the

saving in this case is about 14.55%. The

HVAC modeling parameters and the

customer preference is provided in Table 1.

Table 1. The parameters used in the simulation of

the illustrative example

System

parameters

Customer

preferences

α (0C/kWh) 2.5 tmin (0C) 22

β (p.u.) 0.015 tmax (0C) 26

min

hvacq (kWh) 0

max

hvacq (kWh) 50

CONCLUSION

This paper has presented a new framework

for the energy efficiency of household

customer based on Smartgrid technologies

applied into the existing power grid. The

saving can be achieved by customers actively

responding to the market price which is

passed to the end-users though load

aggregators. The proposed framework is

comprised of two main functions: (1) Off-line

scheduling according to the anticipated data

and (2) On-line control based on both the

ahead scheduling and the real-time

measurements. The problem is formulated

and solved by DP backward algorithm, i.e.,

minimizing the expected energy cost over a

day subject to the uncertainty of the

forecasting. The proposed framework has

been specified and tested in HVAC loads

under real-time electricity. The electricity

price is referred from the PJM electricity

market and the temperature is from National

Climate Data Center in New York, USA in

the same period: from March 24th to 30th 2014

(the whole week). The simulation results

showed that the proposed scheme is not only

capable of controlling the indoor temperature

within the comfortable range (22–260C) set

by customers but the energy costs can be

saved remarkably. The amount of saving over

the simulated period compared with the

traditional operation scheme is as high as

14.55% in this study.


20

REFERENCES

1. A. Brooks, E. Lu, D. Reicher, C. Spirakis and

B. Weihl, “Demand dispatch: Using real-time

control of demand to help balance generation and

load,” IEEE Power and Energy Magazine (2010).

Vol. 5, No. 3, pp. 20-29.

2. S. Majumdar, D. Chattopadhyay and J. Parikh,

“Interruptible load management using optimal

power flow analysis,” IEEE Trans. Power Systems

(1996) Vol. 11, No. 2, pp. 715-720.

3. M.A. Pedrasa, T. D. Spooner and I. F.

MacGill, “Scheduling of demand side resources

using binary particle swarm optimization,” IEEE

Trans. Power Systems (2009), Vol. 24, No. 3, pp.

1173-1181.

4. Z. Fan, “A distributed demand response

algorithm and its application to PHEV charging in

smart grid,” IEEE Trans. Smart Grid (2012), Vol.

3, No. 3, pp. 1280-1291.

5. T. M. Calloway and C. W. Brice, “Physically

based model of demand with applications to load

management assessment and load forecasting,”

IEEE Trans. Power Systems, PAS Vol. 101,

No.12, pp. 4625-4631.

6. J. H. Yoon, R. Baldich and A.Novoselac,

“Dynamic demand response controller based on

real-time retail price for residential buildings,”

IEEE Trans. Smart Grid (2014), Vol. 5, No. 1, pp.

121-129.

7. D. P. Bertsekas, Dynamic Programming and

Optimal Control. Athena Scientific: Belmont,

MA, USA, 1995.

8. PJM Electricity market. Available online:

http://www.pjm.com (accessed on 1st April 2014.)

9. M. Y. Nguyen, V. T. Nguyen, and Y. T. Yoon,

“A new battery energy storage charging/

discharging scheme for wind power producers in

real-time markets,” Energies (2012), Vol. 5, No.

12, pp. 5439-5452.

10. National Climate Data Center (NCDC).

Available online: HTTP://NCDC.NOAA.GOV

(accessed on 1st April 2014).

TÓM TẮT MỘT PHƯƠNG PHÁP TIẾP CẬN MỚI CHO VIỆC TIẾT KIỆM ĐIỆN NĂNG CHO CÁC HỘ TIÊU THỤ DỰA TRÊN CÔNG NGHỆ SMARTGRID

Nguyễn Minh Ý*, Phạm Ngọc Thăng và Nguyễn Huy Toán


Bài báo đề cập đến vấn đề nâng cao hiệu quả trong việc sử dụng điện năng tại các hộ tiêu thụ điện

bằng cách dự báo, cập nhật và xử lý các thông tin về lưới điện; trong thị trường điện, các thông tin

này được phản ánh thông qua giá điện và được truyền đến người dung điện theo thời gian thực

thông qua các công ty bán điện. Trên cơ sở đó, bài báo đề xuất một mô hình quản lý và điều khiển

các thiết bị điện trong hộ gia đình dựa trên những công nghệ của mạng điện thông minh

(Smartgrid) với hàm mục tiêu là tối giảm hóa chi phí tiêu thụ điện năng trong ngày. Bài toán được

mô hình bằng phương pháp quy hoạch động (Dynamic programming) và giải bằng thuật toán tính

ngược (Backward algorithm). Ứng dụng lý thuyết sắc xuất thống kê, các đại lượng ngẫu nhiên như

nhiệt độ môi trường hay nhu của cầu người sử dụng cũng sẽ được giải quyết.

Từ khóa: Điều khiển phụ tải, hiệu suất sử dụng năng lượng, hệ thống điều hòa trung tâm, quy

hoạch động, mạng điện thông minh.


Dinh Thi Gia et al Journal of SCIENCE and TECHNOLOGY 127(13): 21 - 28

21

DIRECT MRAS WITH SAFE CONSTRAINTS APPLIED FOR TWO-WHEELED MOBILE ROBOT

Dinh Thi Gia, Tuan Manh Tran, Son Que Tran*


ABSTRACT

Most two-wheeled mobile robots (TWMR) are controlled and moved by two DC motors. The

heading angular velocity depends on the changing velocity of two wheels mounted on the two DC

motors respectively. During moving, if the heading angular velocity and the linear velocity are too

high, it can lead to flip and slide phenomena. In addition, under the effects of noises (internal and

external), TWMR may be unstable. To solve these problems, we use Euler-Lagrange method to

model for TWMR, build safe conditions against flip, then apply the Model Reference Adaptive

System (MRAS) to construct an adaptive controller for TWMR to ensure the required motion,

stability and safety. Simulation results and analysis point out the effectiveness of the designed

controller.

Keywords: Direct MRAS, Two wheeled mobile robot.

INTRODUCTION *

Two-wheeled mobile robot is shown in Fig. 1

including two wheels, a chassis and a

pendulum. In fact, TWMR - a nonlinear,

unstable and underactuated system - is built

based on the principle of the inverted

pendulum dynamics. To model TWMR, two

widely used methods are Newton and Euler-

Lagrange [1]. With this configuration, It has

been considered as anuseful prototype for

representing nonlinear systems when testing

control algorithms.

To design control for TWMR, the moments

which put into two wheels to control

movement and stability are computed.

When designing controller, the following

parameters are interested: the title angle is

stable at the reference and there is no

overturn while TWMR moving. Although

the system is unstable, difficult to control,

the TWMR is usually used because of the

ability to move in tight space, various terrain

and sharp corners [2].

After linearization, ignoring nonlinear,

coupling attributes, the linear algorithms as

PID, MRAS, etc are applied because they are


quite simple, quick converge, and have small

area stability. On the other hand, the

nonlinear algorithms are complexity, huge

computation, and long response time, but they

have the larger area of stability. However,

under the effect of disturbance, most

conventional controllers can not warrant the

robust performance of system. Normally,

adaptive controller would be the best choice

for this case. It can be easily seen that when

TWMR is drived by human, the TWMR is

affected by unknown forces or disturbance.

This domination is one first daresay in safe

control. The second task must be concerned

that the suitable controller must guarantee that

there are no overturn happening with human

and TWMR. It is quite scarce to find a

controller solving with human safety accept

for author in [2]. In this publication, the

author used a reduced-order disturbance

observer to estimate the disturbance acting on

the TWMR. This estimates disturbance and

compensates in the controller to reduce the

error signal.

In this paper, the model of TWMR is

expanded to three dimensions by using

unequal torque acting on each wheel of

TWMR. It clearly seen that TWMR will


22

rotate itself around z-axis. We continue use

the direct MRAS method designing control

for TWMR to reject system disturbances and

interest to the human safety issues. Two main

issues affecting human feeling are vibrations

when driving under disturbances and safety

when turning with high rotational speed. The

details are presented in Section IV. The

organization of this paper is as follows: After

Section I, introduction, Section II presents the

dynamics of TWMR based on Lagrange

method. In Section III, some basic steps for

designing an adaptive controller based on

direct Model Reference Adaptive System

(MRAS) are presented. Section IV is

mentioned above. In this Section, a MRAS

controller to stabilize position, velocity and

some simulation results are shown. Summary

of this paper is expressed in Section V.

TWO-WHEELED MOBILE ROBOT

In order to design a controller for TWMR, a

dynamic model is first considered. The

equations of motion of TWMR are

established based on balancing forces and

moments on the left wheel, right wheel, the

chassis, and the pendulum. A diagram of

forces and moments acting on the TWMR is

shown in Fig. 1. Definition of parameters and

variables of the TWMR is given in Table 1.

Figure 1. TWMR parameters

Table 1. TWMR parameters and variables

FL, FR Interacting forces between the left

and right wheels and the chassis

HL, HR Friction forces acting on the left and

right wheels

TL, TR Torques provided by wheel actuators

acting on the left and right wheels

fdL, fdR External forces acting on the left and

right wheels

Td External torque acting on the

pendulum

L, R Rotational angles of the left and right

wheels

xL, xR Displacements of the left and right

wheels along the x-axis

Tilt angle of the pendulum

Heading angle of the vehicle

x Displacement of the vehicle along the

x-axis

Mw Mass of the wheel

Jw Moment of inertia of the wheel with

respect to the y-axis

R Radius of the wheel

M Mass of the pendulum

G Gravity acceleration

L Distance from the point O to the

center of gravity, CG, of the

pendulum

D Distance between the left and right

wheels along the y-axis

M Mass of the chassis

Jc Moment of inertia of the chassis

about the y-axis

Jv Moment of inertia of the chassis and

pendulum about the z-axis

Lx , Rx and are chosen as generalized

coordinates. Calculating the kinetic energy for

each component, the final nonlinear dynamics

of TWMR are given as following

2 2

0

2

0

2

0

2

0

( )

sin( )

(

sin( )

sin( )(1)

)sin(

( )

2

)

L RdR dL

x

L RdR dL d

L RdL dR

gm l

J T Tm f f

R R

cosx

l

l cos

l

M mgl m

m T Tf f T

co

R R

D T Tf f

J R

s

R

(1)


23

where

2

ww 22

v

D JJ M J

R

; 2

p cJ m J Jl ;

ww 2

2x

JM M M m

R

2

ww 22

v

D JJ M J

R

; 2

0

2 2 ( )x l cosM J m

In order to make the design controller

simpler, the nonlinear dynamics of TWMR is

linearized. Assuming that TWMR is

symmetrical, the desired balancing angle is

set to zero. The dynamics (1) are linearized

by assuming that with a small angle , the

following equations can be used as

cos( ) 1; sin( ) . It can be seen that the

change of tilt angle affects directly to the

displacement of the vehicle along the x-axis.

When the tilt angle stabilizes to zero, there

also has no displacement along the x-axis.

Ignoring the coupling terms in (1), the

independent linear equations of motion are

rewritten as following

0

1

(2)

2

L Rx dR dL d

L RdL dR

T TM mgl m f f T

R R

D T Tf f

J R R

l

With the disturbance system, the disturbance

usually is rejected by designing a disturbance

observer or using an adaptive controller to

adjust or compensate the annoyance acting on

it. In this paper, an adaptive controller will be

designed as such the disturbance will be

ignored in the dynamics and will be

compensated later. The state space

representation of the system is obtained

0

0

0

00 1

00

(3)

02

02

L

x

R

L

R

ml

R TM mgl

ml T

R

D

RJ T

D T

RJ

PRELIMINARIES

In this section, some basic steps for designing

an adaptive controller based on direct MRAS

method are presented. The general idea

behind MRAS is to create a close loop

controller with parameters that can be updated

to change the response of the system. The

output of the system is compared to a desired

response from a reference model to create an

error signal. This error signal is used to

update to the control parameters so that these

parameters converge to ideal values causing

the plant response to match the response of

the reference model [3, 4].

The structure depicted in Fig 2 can be used as

an adaptive PD controlled system. A second-

order process is controlled with the aid of a

PD-controller. The parameters of this

controller are pK and dK . Variations in the

process parameters pb and pa can be

compensated for variations in pK and dK . We

are going to find the form of the adjustment

laws for pK and dK . The following steps are

thus necessary to design an adaptive

controller with the method of Lyapunov [5]:

Step 1: Determine the differential equation

for e

The description of the process is:

1 2

2 1 2. ( . ) .

p p

p p p p p p d p p p

x x

x b K x a b K x b K

(4)

Aid the state variables and 2 px , where

1pR x (5)

The process in Fig 2 can be described in state

variables:

p p p px A x B u (6)

where

2

p

p

xx

;0 1

. ( . )p

p p p p d

Ab K a b K

;

0

0pb

The desired performance of the complete

feedback system is described by the transfer

function: 2

1

2 22

m n

n n

x

R s z s

(7)


24

By the same way, the description of the

reference model is:

m m m mx A x B u (8)

where

2

m

m

m

xx

;2

0 1

2m

n n

Az

;

2

0m

n

b

and , 1 px , 2 px , R , m , 1 ,mx 2mx , pK , u , m ,

z , and dK are defined in Fig 2.

Subtracting from (8) yields

1 2 1 1 1 2 2 2, ,

m p

m p m m m p p p

T

m p m p

e x x

e x x A x B u A x B u

e e e e x x e x x

(9)

Step 2: Choose a Lyapunov function ( )V e

Simple adaptive laws are found when we use

the Lyapunov function

( ) T T TV e e Pe a a b b (10)

where P is an arbitrary definite positive

symmetrical matrix; a and b are vectors

which contain the non-zero element of the A

and B matrices; and are diagonal

matrices with positive elements which

determine the speed of adaptation.

Step 3: Determine the conditions under which

( )V e is definite negative

2 2

( ) ( )

2 2

T T T T

T T

m p m p

T T

V e Pe e Pe a a b b

A e Ax Bu Pe e P A e Ax Bu

a a b b

(11)

Let: T

m mA P PA Q

where Q is a definite positive matrix.

After some mathematical manipulations, this

yields [6]:

21 1 22 2

11

21 1 22 2 2

22

1( ) (0)

1( ) (0)

p p

p

d p d

p

K P e P e dt Kb

K P e P e x dt Kb

(12)

Step 4: Solve P from T

m mA P PA Q

Figure 2. Adaptive system designed with

Lyapunov

DESIGN CONTROL SYSTEM

In the semi-autonomous TWMR, the change

of head angular velocity is adjusted by human

whereas in the autonomous TWMR, this

value can be set as reference. It can be seen

that when the head angular velocity equal

zero, the head angle should be controlled such

that there is no change under disturbance

while keep tracking the tilt reference angle.

Otherwise, the safe condition relating to the

angular and linear velocity of TWMR is

concerned.

Noting that to stabilize the tilt angle of

TWMR, the tilt angle or displacement x

can be chosen. In this paper, the tilt angle is

selected to warrant reducing the vibration of

pendulum under the effect of disturbance.

This choice also applies for the moving task.

Safe constrain

To clarify the safe condition, forces, acting on

TWMR when TWMR makes the left turning,

are shown in the Fig. 3. Where: w is total

weight, cF is centrifugal force. It is clear that

flipping happens when torque caused by force

cF is greater than torque produced by force

w. Therefore, the safe condition could be

expressed as following

. . / 2.x g D h (13)

Where: g is gravitational acceleration.


25

Figure 3. Forces acting on the TWMR

Adaptive controller based on direct MRAS

In this section, the chosen reference model is

described by the transfer function in (7). The

parameters of reference model are selected

such that the higher order dynamics of the

system will not be excited. With these

concern, 10[ / ]n rad s and 0.7z are picked.

As such:

2

0 1 0 1

2 100 14m

n n

Az

;

0

100mb

The adaptive control gains, described in

integral form, are described as follows

21 1 22 2

11

21 1 22 2 2

22

1( ) (0)

1( ) (0)

p p

p

d p d

p

K P e P e dt Kb

K P e P e dt Kb

(14)

21 1 22 2

11

21 1 22 2 2

22

1( ) (0)

1( ) (0)

p p

p

d p d

p

K P e P e dt Kb

K P e P e dt Kb

(15)

21 1 22 2

11

21 1 22 2

22

1( ) (0)

1( ) (0)

p p

p

i i

p

K P e P e dt Kb

K P e P e dt Kb

(16)

In the form of the adjustment laws,

21( )P and 22( )P are elements of ( )P matrices,

obtained from the solution of (17).

( ) ( ) ( ) ( ) ( )

T

m mA P P A Q (17)

where ( ) stands for , , and ,

respectively.

Simulation results

The parameters for simulation are listed as

follows: Mw = 1 kg, Jw = 1.5 (kgm2), R =

0.25 (m), m = 1.5 (kg), g = 9.81 (m/s2), l = 1.2

(m), D = 0.15 (m), M = 5 (kg), Jc = 2.5 (kgm2)

and Jv = 1.21875 (kgm2). The simulation

control structures are presented in two cases.

In Fig. 4, the reference inputs are chosen such

that d is equal zero for stabilization. Fig.5 is

control structure of TWMR with safe

condition in which is not equal zero.

The adaptive controller parameters are

selected by setting:

( ) 0.03p , ( ) 3d , ( ) 3i ,

( )Q = [350 100; 100 250],

( ) 10[ / ]n rad s and ( ) 0.7z

Fig. 6 illustrates the responses of TWMR for

the stabilization case. It can be seen that tilt

angle and head angle converge origin

whereas the displacement of TWMR x is not

change when TWMR is stable. The adaptive

control gains are shown in Fig. 7. Fig 8

presents the response the linear velocity and

reference angular velocity with safe constrain.

It is clear that under safe condition, d

adjusts to a suitable value such that it satisfies

the safe condition. However, the tilt angle

keeps stable under disturbance.

CONCLUSION

Safe controller is invested in two cases:

TWMR must be stable under disturbance and

satisfy the safe condition. Based on direct

MRAS method, an adaptive controller has

been completed. The simulation results show

that the designed controller fulfill control

objective. Under disturbance, TWMR keeps

stable and almost has no vibration on the

pendulum. Moreover, the safe condition

warrants having no overturn and can directly

apply for semi-autoromous control.


26

e_Phi

e_Psi

TLPlusTR

TLMinusTR

Phi

Phi_dot

Psi_dot

Psi

TWMR

0

Psi_d

0

Phi_d

Ref

Out

Out_dot

PD Model 2

Ref

Out_dot

Out

PD Model 1

In

In_dot

e_Psi

In_m

In_m_dot

TLMinusTR

MRAS PD Psi

In_m_dot

In_m

e_Phi

In_dot

In

TLPlusTR

MRAS PD Phi

Signal 1

Disturbance 2

Signal 2

Disturbance 1

Figure 4. Simulation control structure

e_Psi_dot

e_Phi

x_dot

psi_dot

psi_dot_d

SafeConstrain

.3

Psi_dot_d

TLPlusTR

TLMinusTR

Phi_dot

Phi

Psi_dot

Psi_2dot

x_dot

Plant

0.15

Phi_d

Ref

Out

Out_dot

PI Model 2

Ref

Out_dot

Out

PD Model 1

In_dot

In

e_Psi_dot

In_m

In_m_dot

TLMinusTR

MRAC PI Psi_dot

In_m_dot

In_m

e_Phi

In

In_dot

TLPlusPR

MRAC PD Phi

Signal 1

Disturbance 2

Signal 1

Disturbance 1

Figure 5. Simulation control structure


27

Figure 6. Responses of the adaptive control


Figure 7. Adaptive gains

Figure 8. Reference angular and linear velocity

after using the safe constrains

Figure 9. Adaptive gains

Figure 10. Responses of the adaptive control



28

REFERENCE

1. R. P. M. Chan, K. A. Stol, and C. R. Halkyard,

"Review of modelling and control of two-wheeled

robots," Elsevier, vol. 37, 2013.

2. D. Choi and J.-H. Oh, "Humand-friendly

motion control of a wheeled inverted pendulum by

reduced-order disturbance observer," presented at

the 2008 IEEE International Conference on

Robotics and Automation, Pasadena, CA, USA,

2008.

3. N. D. Cuong, G. T. Dinh, and T. X. Minh,

"Direct MRAS based an Adaptive Control System

for a Two-Wheel Mobile Robot," in 2014 2nd

International Conference on Control, Robotics and

Cybernetics (ICCRC2014), Singapore, 2014.

4. J. V. Amerongen, "Intelligent Control (part 1) -

MRAS," ed University of Twente, The

Netherlands, 2004.

5. C. J. Kaufman, "Boulder, CO, private

communication," Rocky Mountain Research Lab,

May 1995.

6. Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa,

"Electron spectroscopy studies on magneto-optical

media and plastic substrate interfaces (Translation

Journals style)," presented at the Dig. 9th Annu.

Conf. Magnetics Japan, Aug. 1987.

TÓM TẮT

MRAS TRỰC TIẾP VỚI CÁC ĐIỀU KIỆN AN TOÀN ỨNG DỤNG

ĐIỀU KHIỂN ROBOT DI ĐỘNG HAI BÁNH

Gia Thị Định, Trần Mạnh Tuấn, Trần Quế Sơn*


Hầu hết các Robot di động hai bánh (TWMR) được điều khiển và di chuyển bởi hai động cơ DC.

Vận tốc góc đỉnh phụ thuộc vào sự thay đổi vận tốc của hai bánh xe có gắn hai động cơ DC. Trong

khi di chuyển, nếu vận tốc góc đỉnh và vận tốc dịch chuyển của xe quá lớn thì có thể dẫn tới hiện

tượng robot bị lật và trượt, thêm vào đó là sự tác động của nhiễu (nhiễu trong và nhiễu ngoài) làm

TWMR mất ổn định. Để giải quyết các vấn đề trên, chúng tôi đã sử dụng phương pháp Euler-

Lagrang để mô hình hóa TWMR, xây dựng các bộ điều khiển ứng dụng điều khiển thích nghi theo

mô hình mẫu trực tiếp có kể đến các điều kiện an toàn để đảm bảo rằng TWMR chuyển động, ổn

định và an toàn. Từ đó chỉ ra những hiệu quả của thiết kế trong việc điều khiển TWMR thông qua

các phân tích cũng như các kết quả mô phỏng.

Từ khóa: Hệ thống thích nghi mô hình mẫu trực tiếp, Rô bốt di động hai bánh xe.


Phong Tien Le et al Journal of SCIENCE and TECHNOLOGY 127(13): 29 - 35

29

SEARCH ON DESIGNING AN ENERGY MANAGEMENT SYSTEM

FOR ISOLATED PHOTOVOLTAIC SOURCE

Phong Tien Le*, Minh Duc Ngo


ABSTRACT

Energy receiving from PV (Photovoltaic) source depends on the solar irradiance also exploited

method, quality of PV. This paper presents an isolated PV system that can exploit maximum

energy from PV, protect system automatically whenever having an requirement in breaking down

circuit or having any faults to ensure electric power for load. It also proposes an automatic energy

management system to control power operating point of PV, capacity charging and discharging for

battery. This system is built and tested in an experiment model: PV-battery-load system that can

adapt to an existed PV source to supply electric power for LED lighting.

Keywords: Solar cell, PV, maximum power point, P&O, energy management system.

INTRODUCTION

PV is considered as one of the most potential

sources in renewable energy to ensure long

lives on the earth. It has many orientations in

theory and the experiment of researchers in

the world focuses on enhancing electric

quality or the ability of exploiting such as

finding maximum power point (MPP) of PV,

improving converters or efficiency of PV,etc.

[1-3]

PV power depends on the solar irradiance and

temperature. Generally, it increases in the

morning, reaches the maximum at noon,

decreases in the afternoon and vanishes

completely at night. Because of low

efficiency and the technology for producing

PV far from expectation, energy experts have

to find other methods to enhance the ability of

exploiting PV power.

PV power can be used in two configurations:

isolated grid (for local loads) or connected to

the main grid (as distributed source) in

distribution system. This paper designs a

model: PV-battery load, that can charge

energy to a battery bank, protect components

based on a hardware architect and build an

energy management system.

CHARACTERISTIC OF PV SOURCE

V-I characteristic

The current generated from PV is the flow of

electrons made by irradiance on

semiconductors. The relationship between

voltage and current of a PV module is shown

by V-I characteristic in Fig. 1.

0 5 10 15 20 250

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Voltage [V]

Cur

rent

[A

]

2

800 W/m

600 W/m

400 W/m

1000 W/m

2

2

2

0 5 10 15 20 25

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Voltage [V]

Curr

ent

[A]

20 C

40 C

60 C

0

0

0

a. Change irradiance (Constant temperature) * b. Change temperature (Constant irradiance)

Figure 1. V-I characteristic of a PV module



30

The V-I curve tends to be higher (Fig. 1a)

when the irradiance increases from 0 (early

morning) to maximum value (approximately

1000W/m2 at noon) and be lower in the

afternoon.

The V-I curve tends to move left and open

voltage decreases (VOC) when it increase

temperature (Fig. 1b).

We can see that the variation of irradiance

affects the V-I characteristic more than the

variation of temperature. Moreover, PV

voltage often have value in fixed range. At

the time of low irradiance, VOC can be

measured and has a value but current is so

small that it cannot remain control system.

Maximum power point (MPP)

When irradiance or temperature changes, V-I

curve also changes but there always exists a

point of maximum power (Fig. 2). To operate

at this point, the control system has to set

input voltage of control circuit at the voltage

of MPP.

0 5 10 15 20 250

20

40

60

80

100

120

Voltage [V]

Pow

er

[W]

1000W/m2

700W/m2

400W/m2

MPP1

MPP2

MPP3

Figure 2. P-V characteristic

ENERGY MANAGEMENT SYTEM FOR

AN ISOLATED PV SOURCE

The structure of system

The structure of isolated PV systems is shown

in Fig. 3

The energy from PV is transfomed through

DC/DC converter to reduce voltage to

(15÷17)V that is suitable to charge battery.

The current and voltage signals generating

from PV are collected, converted to digital

signal and sent to central processor to track

the variation of energy at each time. It also

controls the input voltage of DC/DC

converter by changing the value of PWM

(Pulse Width Modulation). The current and

voltage signals of output DC/DC converter

are collected continuously to observe, and

track the capacity of battery.

DC/DC converer

DC/DC converter type flyback is the simplest

converter because it has only one switch, one

transformer and no inductor at output. The

transformer isolates PV (primary) and output

(secondary); it also has low cost and adapts to

the variation of source faster than other

converters. It uses high frequency to have a

smaller transformer, filter inductors, and

capacitor, and achieve a faster dynamic

response to rapid changes in the load current

and/or the input voltage. DC/DC type flyback

topology the principle of this converter and is

shown in Fig. 4 [7].

Figure 3. The structure of isolated PV system

PV Solar

irradiance

DC/DC

Converter

Energy

management

system

Energy

storage

Timer

Low voltage

relay

Load Display block

i

u

Control

signal

u

i


31

Figure 4. Operation modes of flyback converter

Energy is stored in the primary winding when

the switch S is on and energy is transformed

to the secondary winding when S is off. It can

be separated in three case:

Case 1: S is turned on, diode D is turn off, the

voltage across the primary winding is Vp, the

current through it increases linearly from

zero.

Case 2: S becomes turned off, D turns on, the

voltage across the primary winding becomes

(−Vp). Therefore, the current through the

primary winding decreases linearly. This

current is reflected in the secondary winding

of the transformer and flows through D.

Case 3: S and D is off, no energy is

transformed through this converter.

Energy management system

● System tests the ability of supplying energy

for converter (Fig. 5)

For PV source, the lower irradiance is, the

lower V-I characteristic. Specially, when it is

nearly dark, the energy supplying for

converter decreases faster and if it has no

system for testing the ability of supplying

from PV source, the process of closing and

opening circuit happens continously causing

damage to components.

Figure 5. Test for the ability of supplying from PV

source and the protection of circuit

Charging algorithm for battery

When the energy storage is battery and

charging time is (810)h/day corresponding

to (7580)% capacity of battery C0, we can

charge in constant current at 10%C0 (Mode

1). After reaching to this capcity, the battery

starts boiling and goes on constant voltage

charging to have full battery (Mode 2). When

IPV < Iref, energy management system will

activate P&O algorithm to find MPP. They

are shown in Fig. 6 and Fig. 7.

PV

Panel

S *

*

D

C1 C2

a. DC/DC converter type flyback

Batter

y

PV

Panel

S *

*

D

C1 C2

B

ip iS

b. Case 1: S is on, D is off

+

Vp

- +

- VS

PV

Panel

S *

*

D

C1 C2

B

ip iS

c. Case 2: S is off, D is on

PV

Panel

S *

C1

ip

*

D

C2

B

iS

d. Case 3: S is off, D is off

Vp VS

Vp VS

+

+

+

+

-

-

-

-

Start

Open circuit

Remain time t

N

Charging

algorithm

Stop

Y

PPV > Pcharging


32

P&O method is easy to use because it only

needs to change pulse control to find MPP.

Power at the present is compared with the

previous power: if PV voltage changes and

power increases (dP/dV>0), the control

system will move operating point to another

point for increasing power; it will move to

opposite direction if dP/dV<0. Although PV

power is fluctuated around MPP and it causes

oscillation of power in the system but it can

be accepted because we have to consider that

there is inertia of device in control circuit

(capacitors, inductors) added to the oscillation

of power [6].

DESIGN A PV EXPERIMENT SYSTEM

PV parameters

The PV modules have been located in Thai

Nguyen Unversity of Technology since 2009

by Phoenix Solar Pte, Singapore. PV

parameters are shown in table 1.

Test PV capicity

Energy E(Wp) adapting to load per day can be

caculated by equation (1):

0. .

( ) (1). ( ).

loadP

S m med

P t EE W

T E

where

total load is Pload = 4.40 W (4 LED luminares)

time to adapt for load is t = 9 hours per day

the efficiency of PV source is s = 0,6 the

correction factor for PV capacity by

temperature is m(T) = 0,7 the average

irradiance energy is approximately:

Emed = 3000 Wh/(m2.day)

(Thai Nguyen’s location is Northern area) the

standard irradiance is E0 = 3000 Wh/m2.

Therefore E(Wp) = 1142,9Wp

Caculating in theory, it can adapt 119%

power requirement in the normal operating at

MPP.

Table 1. Parameter of PV source

Name of

parameter

Type: Kyocera

KC85/Japan

Open voltage

(VOC)

Short circuit

current (ISC)

PV power at

MPP (Pmax)

Votage at MPP

(VMPP)

Current at

MPP (IMPP)

Value 16 module in

series 21,6 V 4,9 A 16 x 85 Wp 17,3 V 4 A

Decrease V(k) (Increase PWM)

Increase V(k) (Decrease PWM)

Y

N

V(k)>V(k-1)

V(k)<V(k-1)

N

Y

N

Y Y

Figure 7. P&O algorithm

Start

Set , t0

P(k)–P(k-1)>

Decrease V(k) (Increase PWM)

Increase V(k) (Decrease PWM)

N

P(k)–P(k-1)≤

Measure V(k), I(k)

t ≤t0 N

Icharging= Imax

Y

Figure 6. Charging algorithm for battery

Scan PWM

Icharging=Iref

P&O algorithm

N

Y

Icharging

Start

Battery is full Y

Stop N

Test Iref

Icharging ≥ Iref

Mode 1

Y

Mode 2 N

Vcharging


33

Caculate the capacity of battery

The total capacity of battery can be caculated

by equation (2):

..

.

DODV

DEC

ng

(A.h) (2)

where, V is the average voltage of battery

(choose V = 12V), is the efficience of

battery (choose = 0,95), DOD is the depth

of discharging (choose DOD = 0,8), D is the

backup factor for weak irradiance day (in

winter or cloudy), choose D = 1,2.

Choose the capacity of battery is 200Ah.

Caculate parameters for converter

Switching Frequency of converter switches:

fsc =200kHz.

Turns ratio for flyback tranformer can be

caculated by equation (3)

max2

1 max 0

(3)(1 )

inD VNn

N D V

Vin = (150300)V, V0 = (1517)V. Choose

n=10/1

Output capacitor can be caculated by

equation (4)

max 02 (4)

sc L cpp

D VC

f R V

Vcpp =1V (allowed ripple voltage across the

output capacitor). Choose C = 4000F.

The ability of operating system

Operating mode: having irradiance, the

battery operates in charging mode. Without

irradiance, the battery operates in discharging

mode and supplies power for load.

Control circuit can maintain charging

current for battery although PV output voltage

is in fixed range.

Cut down load whenever votage of battery

reduces to a low value to protect battery.

Moreover, it can open or close load at fixed

time by timer.

Reduce the fluctuation of power for circuit

at the time having low irradiance (nearly

dark) to protect components by a time relay.

EXPERIMENT RESULTS AND

CONCLUSIONS

Experiment results

● Time case 1: From 4.00pm to 4.13pm on 31

July 2014 and Iref=15A. It’s sunny, sunlight

falls down PV panel directly. Tracking PV

system and Icharging diagram are shown in Fig.

8 and Fig. 9.

Figure 8. Tracking PV system in case 1

Figure 9. Icharging diagram in case 1

● Time case 2: From 7.17am to 7.28am on 1

August 2014 and Iref=7A. Sunlight doesn’t

fall down PV panel directly. Tracking PV

system and Icharging diagram are shown in Fig.

10 and Fig.11.



34


● Time case 3: From 9.28am to 9.38am on 1

August 2014 and Iref = 20A. PV panel is

eperated in partially shaded condition.

Tracking PV system and Icharging diagram are

shown in Fig. 12 and Fig. 13.



Table 2. Efficiency of system

Case Time [%]

1 4.00pm on 31 July 2014 61,6

2 7.17am on 1 August 2014 51,8

7:28am on 1 August 2014 59,61

3 9.28am on 1 August 2014 64,1

Conclusions

Input power always fluctuates when PV

power is insufficient and the processor

changes duty cycle D to find other values

nearer Iref by using P&O algorithm whenever

Icharging < Iref. It shows that the control system

tracks Iref very well.

The efficiency of systems changes in range.

It depends on Iref and other factors of system:

the capacity of battery (full, near full, empty,

etc.), PV capacity and own charging circuit.

Charging power curves are often smoother

than PV output power curves because of the

output capacitor of converter. Output voltage

of converter is changed by Iref.

Energy management system administrates

well using algorithms to exploit PV power

best in an isolated PV system. It can be

enlarged for connecting system to the grid.

REFFERENCE

1. Hoang Duong Hung, Solar energy – Theory and

application, Science and technology publisher,

2007.

2. Dang Dinh Thong, Le Danh Lien, Basic new

and renewable energy, Science and technology

publisher, 2006.

3. Than Ngoc Hoan, Photovoltaic Energy and

Methods Increasing his Quality and Efficiency,

Journal of the Marine Science and Technology,

No.18, 2009.

4. Joe-Air Jiang, Tsong-Liang Huang, Ying-Tung

Hsiao, Chia-Hong Chen, Maximum power

tracking for photovoltaic power system, Tamkang

Journal of Science and Engineering, Vol. 8, No 2,

pp. 147-153, 2005

5. T.Chaitanya, Ch.Saibabu, J.Surya Kumari,

Modeling and Simulation of PV Array and its

Performance Enhancement Using MPPT (P&O)

Technique, T.Chaitanya et al, International Journal

of Computer Science & Communication

Networks, Vol 1,September-October 2011.

6. Roberto Faranda, Sonia Leva, Energy

comparison of MPPT techniques for PV Systems,

Wseas Transactions on power systems, ISSN:

1790-5060, Issue 6, Volume 3, June 2008.

7. Neha Adhikari, Bhim Singh, A.L.Vyas, Design

and Control of Small Power Standalone Solar PV

Energy System, Asian Power Electronics Journal,

Vol. 6, No. 1, Oct. 2012.


35

TÓM TẮT

NGHIÊN CỨU THIẾT KẾ HỆ THỐNG QUẢN LÝ NĂNG LƯỢNG

CHO NGUỒN PIN MẶT TRỜI ĐỘC LẬP

Lê Tiên Phong*, Ngô Đức Minh Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên

Năng lượng nhận được từ nguồn PV phụ thuộc vào cường độ bức xạ cường bộ bức xạ từ mặt trời

cũng như phương pháp khai thác, chất lượng của nguồn PV. Bài báo này trình bày một hệ thống

PV cô lập để khai thác năng lượng lớn nhất từ nguồn PV, hệ thống bảo vệ tự động khi có yêu cầu

cắt mạch điện hoặc có bất kỳ sự cố nào để đảm bảo công suất điện cho phụ tải. Bài báo cũng đề

xuất một hệ thống quản lý năng lượng tự động để kiểm soát điểm vận hành cho nguồn PV, khả

năng phóng nạp cho ắc quy. Hệ thống này được xây dựng và kiểm tra cho một mô hình thực

nghiệm: nguồn PV-ắc quy-tải để đáp ứng với một nguồn PV cho trước, qua đó cung cấp điện cho

các đèn LED chiếu sáng.

Từ khóa: Pin mặt trời, điểm làm việc công suất cực đại, P&O, hệ thống quản lý năng lượng.

* Tel: 0986938968Email: [email protected]


36


37

CONTROL OF GRID-CONNECTED SOLAR POWER SYSTEMS WITH

INTERLEAVED FLYBACK CONVERTERS

Phong Tien Le*, Huong Thi Mai Nguyen, Hung Tien Nguyen University of Technology – TNU

ABSTRACT This paper proposes a control system connecting for the grid-connected PV (Photovoltaic) system.

The control structure includes two alternating flyback converters, a control unit for connecting PV

system to the grid, filter circuits at input and output of the converters. The output voltages of two

flyback converters having half sinusoidal wave are is converted to completely sinusoidal wave by

unfolding converter with the grid synchronization. This control system monitors the load demand

and generated power capacity of PV in order to regulate power getting from the grid. The

simulation is carried out bay MATLAB/Simulink.

Keywords: Photovoltaic panel, flyback converter, connecting grid, inverter, regulate power.

INTRODUCTION *

Electric power from PV is only generated in

daylight time having irradiance and stored in

batteries to adapt to the load or generated

directly to grid by electronic power converters

[1-3].

Converters are widely used in many fields

that convert DC to AC using electronic power

components such as IGBT or MOSFET. With

the development of digital microprocessors,

more and more converters have high quality,

simple structure and low cost [1-4].

There are two different kinds of inverters. The

first kind is voltage source inverter and the

second kind is current source inverter in

which the output current is AC with fixed

power factor. For the control of a grid-

connected PV power system, we consider a

PV panel with a maximum power point

tracker (MPPT) to have voltage operating at

fixed value.

A design of a PV power system is presented

in detail in [5]. However, this report was not

completed since it still has some missing

parts. Hence, this paper is a complementary

for [5], in which we will present the whole

controlled system with unfolding converter

and the intergration of a simulated MPPT

block.

*Tel: 0986938968; Email: [email protected]

INTERLEAVED FLYBACK CONVERTER

The DC/DC flyback converter type is the

simplest one since it has only one switch, one

transformer and no inductor at the output.

Another type of the flyback converter is

alternating flyback converters that operate in

discontinuous mode with two flyback

converters as shown in Fig. 1.

Figure 1. Interleaved flyback converter

The principle operation of this converter is

following [6]:

- Step 1 (t0t1): The main switch Q1 is

allowed to conduct and Q2 is non-conducted.

The current ipv1 at the primary side of T1

flows from source to Q1 and backs to source.

Energy is stored in main transformer T1. At

the same time, the energy from capacitor C0 is

supplied to load (Fig. 2).


38

Figure 2. Current in converter at step 1

- Step 2 (t1t2): Both main switches Q1 and

Q2 are non-conducted. The current is1 at the

primary of transformer T1 flows through D1

and C0 to supply to load. Energy stored in the

main transformer T1 is released (Fig. 3).


- Step 3 (t2t3): The main switch Q2 is

allowed to conduct and Q1 is non-conducted.

The current ipv2 at the primary side of

transformer T2 flows from the source to

switch Q2 and backs to source. Energy is

stored in main transformer T2. At the same

time, the energy from capacitor C0 is supplied

to the load (Fig. 4).


- Step 4 (t3t4): Both main switches Q1 và Q2

are non-conducted. The current is2 at the

primary side of the transformer T2 flows

through D2 and C0 to supply to the load.

Energy stored in the main transformer Q2 is

released (Fig. 5)


DESIGN OF CONTROL SYSTEM

The current control loop

Flyback converter is a high non-linear system.

A single non-ideal flyback grid-connected is

shown in Fig. 6.

Figure 6. Non-ideal flyback converter

It is assumed that the inverter voltage has

same amplitude with the grid. This simplifies

the analysis of flyback converter. Note that

the flyback converter has three levels

responding following three stored energy

levels:

im – current in flyback inductor.

vac(s) – voltage in output capacitor.

iac(s) – current in output filtering inductor.

Average current and voltage Kirchohoff

equations of converter in one switching cycle

are shown as:


39

0

. . .( ) '.

. '

. (1)

.

.

m

lm m

ac m s

pv m on p

m

s

lf ac ac f grid

ac

C s ac

grid load ac

pv m

div L

dt

v i Rd v d i R R d

N

ii d

N

v v i R v

dvi C i i

dt

v R i

i d i

where dd is the operating cycle and d’=1-d is

the time interval between two cycles.

All quantities (at levels and inputs) in

equation (1) are average in one switching

circuit. Each equation in (1) is large signal

and describes accuracy (non-linear) system.

In order to develop a transfer function

between output and input control as well as

disturbance input, the system equations are

approximated at a chosen operating point.

Note that the small PV system has a wide

voltage and current range. Because the

converter operates at unity power factor, the

resister R of the load can be calculated by

vgridrms/iacrms. The operating point is chosen

corresponding to the RMS nominal voltage

value and output current at MPPT block.

The state and input vectors are expressed as: T

acacm viix ][ (2)

T

pvgrid vvdu ][ (3)

where d is control input, vgrid and Vpv is input

disturbances.

Since the control variable is current of output

filter, the controlled output of the system is

][ aciy (4)

Changing and isolating state, input and output

vectors it follows as:

yYy

uUu

xXx

~

~

~

acac

pvgridpvgrid

acacmacacm

iI

vvdvvD

viivII

~]~~~

[][

]~~~[][

(5)

where X, U, and Y represent static operating

points, and yvàux ~~,~ are disturbances on the

operating points.

Displace equation (5) to equation (1), split

AC quantities and remove small AC

components we obtain the following linear

system:

( )

'( ) '

'

m

lm m

m ac

pv pv m on p s

s ac

on p m

ac m

s

ac

lf f ac ac f grid

div L

dt

i vDv d v i R R R

N N

R D vD R R D i

N N

D i dii

dt N

div L v i R v

dt

mmpv

acm

mac

c

idiDi

idN

ii

N

D

dt

vdCi

~~~

~~~'~~0

(6)

From equation (6), the small signal AC circuit

is drawn as in Fig. 7.

Figure 7. Small-signal AC circuit

Rewrite equation (6) we have:

pvgridm

acacmac

pvgrid

f

ac

f

ac

f

f

mac

pv

m

grid

m

ac

m

acm

m

m

vvdNC

Ivi

Ci

NC

D

dt

vd

vvL

dvL

iL

Ri

dt

id

vL

Dvd

L

kv

NL

Dii

L

R

dt

id

~.0~.0~~.0

~.

1~'~

~.0~.1~

.0~1~.

~.0

~

~~.0~~'~

.0~

~

000

(7)

where

N

vR

N

IRRIvk ac

sm

ponm

~)(~

N

RDRRDR s

pon ')(

And aciy~

(8)


40

Write equation (7) and (8) in the matrix form:

0 0

0

'0

10

' 10

0

10 0

00

m

m m

m

fac

ac

f f

acac

m m

pv grid

f

m

R DdiL L Ndt i

Rdii

dt L Lv

dv Ddt NC C

k D

L L

d v vL

I

NC

(9)

332211 uBuBuBAxx (10)

0 1 0

m

ac ac

ac

i

i i

v

(11)

y=Cx (12)

The relationship between output and input

controls (all disturbances equal 0) is

represented as follows:

1

1

1

)()(

)(BAsIC

sU

sY (13)

with U2(s) = U3(s) = 0.

Using equation (13), the transfer function

between output AC current and input can be

written as:

2

1

id

idid

G

GG (14)

where

NCL

sI

CNLL

RIkDG

m

m

fm

mid

00

1

'

23 2

2 2

0 0

2

2

0 0

1 '

'

f

id

f m f f m

m

f

f

m f m

RRR DG s s s

R L L L C N L CL

L

R DR

L L C N L C

Design feed-forward compensator

A system combining feedback and feed-

forward can enhance the performance of

controlled system in the presence of

disturbances. In an ideal condition, a feed-

forward compensator will fully remove

disturbance. The role of the feed-forward

compensator in small PV system is to supply

steady-state ratio, “D(t)”, to system and

“∆d(t)” to follow current reference. It also

removes disturbance caused by fluctuations of

the PV and AC grid voltages. The relationship

between input and output voltage of the

flyback is:

inV

D

NDV

10

(15)

Rewrite equation (15) in steady-state cycle D

we have

NVV

VD

in

0

0 (16)

Note that (16) represents the relationship

between output/input voltage at steady-state.

The final operating ratio is expressed as:

d(t) = D(t) + ∆d(t) (17)

where D(t) is contribution from feed-forward

compensator and ∆d(t) is contribution from

the AC control compensator.

Design of load sharing control loop

As above analysis, the small PV system

includes two flyback converters connecting in

output and input parallel (IPOP) to share a

load. For any of the two real converters,

although designed the same, is subjected to

parameter changes such as the secondary

resistance, Rdson, core loss… This might

cause overload in one converter. Therefore, a

combination of sharing compensator is highly

important to obtain equal shared currents.

The control loop always supervises variation

between input currents of the converters and

corrects the operating ratio of each converter

by adding or subtracting a small corrective


41

ratio depending on the sign of the variation.

The transfer function current I(s) and the

accurate corrective factor d(s) are:

Ipv1 = Gd,ipv1(s)Xd(s) (19)

Ipv2 = Gd,ipv1(s)Xd(s) (20)

Variation between currents is I, hence the

currents can be seen as:

I=Ipv1(s)-Ipv2(s)=(Gd,ipv1(s)-Gd,ipv2(s))Xd(s) (21)

The shared currents include two equal parts

(Ipv1=Ipv1-I/2 and Ipv2=Ipv2+I/2) with the

corrective factor ±d.

Rewrite equations (19) and (20) with the

corrective factor we obtain:

Ipv1(s) - I/2= Gd,ipv1(s)X(d(s) - d(s)) (22)

Ipv2(s)+I/2= Gd,ipv2(s)X(d(s) + d(s)) (23)

Subtract equation (23) to equation (22) we

have:

I = (Gd,ipv2(s) - Gd,ipv1(s))d(s) + (Gd,pv2(s) +

Gd,ipv1)d(s) (24) Assume Gd,ipv2 Gd,ipv1 Gd,ipv

then: I(s) = 2Gd,ipv(s)d(s) (25)

SIMULATION RESULTS

The Simulink model of the controlled system

is shown in Fig. 8.

Figure 8. Diagram of system simulation

In order to see the adapbility of the controlled

system, the PV operation point at MPPT is

changed from (36V, 215W) to (40V, 350W)

at the time 0.08s. The output voltage and

current of inverter are shown in Fig. 9 and 10.

As can be seen in Fig. 9 and Fig. 10, The

output voltage magnitude does not change

although the PV voltage increase from 36V to

40V, while the output current only change its

magnitude. It is significant to connect the grid.

0.05 0.1 0.15 0.2

-300

-200

-100

0

100

200

300

The output voltage

time (s)

Vol

Figure 9. Output voltage of the flyback converter

0.05 0.1 0.15 0.2

-2

-1

0

1

2

3

4The output current

time (s)

A

Instantaneous value

RMS value

Figure 10. Output currents of the inverter

CONCLUSIONS

This model depicted a control system to

connect PV source to the grid. It uses

double flyback converters, shared load

control, feed-forward compensator.

Moreover, an unfolding converter

providing an exactly sinusoidal voltage is

proposed in this paper to have a simple

control system.

Simulation results show that output voltage

has a sinusoidal fixed form corresponding

with grid voltage every time although PV

power changes. Controlled system adapt

well to connect PV source to the grid.


42

REFERENCE

1. Neha Adhikari, Bhim Singh, A.L.Vyas, Design

and Control of Small Power Standalone Solar PV

Energy System, Asian Power Electronics Journal,

Vol. 6, No. 1, Oct. 2012.

2. Martina Calais, Johanna Myrzik, Ted Spoone,

and Vassilios G. Agelidis, Inverters for single-

phase grid connected photovoltaic systems-an

overview, IEEE 33rd Annual on Power Electronics

Specialists Conference, 4:1995 – 2000, 2002.

3. Bjornar Gundersen, An investigation on

gridconnectable single phase photovoltaic

inverters, Master’s thesis, Norwegian University

of Science and Technology, 2010.

4. Frede Blaabjerg, Remus Teodorescu, Zhe Chen,

and Marco Liserre, Power converters and control

of renewable energy systems, Aalborg University,

2006.

5. Microchip, Grid-connected solar microinverter

reference design, Technical report, 2012.

6. K.I. Hwu and C.F. Chuang, Development of

interleaved control for the pfc flyback converter,

IEEE Region 10 Conference. TENCON, 2007.

TÓM TẮT

ĐIỀU KHIỂN HỆ THỐNG PIN MẶT TRỜI KẾT NỐI LƯỚI

SỬ DỤNG BỘ BIẾN ĐỔI FLYBACK ĐAN XEN

Lê Tiên Phong*, Nguyễn Thị Mai Hương, Nguyễn Tiến Hưng


Bài báo đề xuất một hệ thống điều khiển hệ thống PV kết nối lưới. Cấu trúc điều khiển bao gồm

hai bộ biến đổi flyback đan xen, bộ điều khiển kết nối hệ thống với lưới, các mạch lọc ở đầu vào

và đầu ra của bộ biến đổi. Các điện áp ra của hai bộ biến đổi flyback có dạng sóng nửa sin được

chuyển thành dạng sin hoàn toàn bằng cách sử dụng bộ lật để đồng bộ với lưới. Hệ thống điều

khiển này kiểm tra yêu cầu của phụ tải và khả năng phát công suất của nguồn PV để điều chỉnh

công suất nhận từ lưới. Mô phỏng hệ thống được thực hiện trên phần mềm Matlab/Simulink.

Từ khóa: Pin mặt trời, bộ biến đổi flyback, kết nối lưới, bộ nghịch lưu, điều chỉnh công suất.


Nam Hoai Nguyen et al Journal of SCIENCE and TECHNOLOGY 127(13): 43 - 46

43

A NEW TRAINING PROCEDURE FOR A CLASS

OF RECURRENT NEURAL NETWORKS

Nam Hoai Nguyen1,*, Nguyet Thi Minh Trinh2 1University of Technology – TNU, 2Yen Bai Collegue of Technique

ABSTRACT This work is to propose a new training procedure for a class of recurrent neural networks. Based

on reservoir computing networks, we extend their network structure from one delay to more than

one delay and modify their training method. The novel training method is demonstrated on a

benchmark problem and an experimental robot arm and compared to traditional training methods.

The result shows that the proposed training procedures give some better advantages such as

smaller number of weights and biases andfaster training time.

Keywords: Recurrent neural networks, reservoir computing network, echo state network, training

procedure, system identification, one link robot arm.

INTRODUCTION*

Reservoir computing networks (RCNs) have

been successfully used for time series

prediction. There are two major types of

RCNs: Liquid-state machines [1] and Echo-

state networks [2]. An input signal is fed into

a fixed weights dynamic network called

reservoir and the dynamics of the reservoir

map the input to the reservoir’s state. Then a

simple readout mechanism is trained to read

the state of the reservoir and map it to the

desired output.

The capability of system identification of

RCNs is limited because of being only first

order models. Thus, RCNs are unable to

identify systems of higher order. But we can

apply the philosophy of RCNs training to

train other types of recurrent neural networks.

Here we focus on the structure of neural

networks given in Fig. 14 of the work [3]. It

can be shown as in Fig. 1. This type of neural

networks is widely used in identification and

control of dynamic nonlinear systems.

In the next section, a new training procedure

is proposed. A structure of recurrent neural

networks is described and a novel training

method is given. The following section is

applications of the proposed training

procedure to systems identification. Two

examples are represented. In the final section,

conclusions and future work are provided.


PROPOSED TRAINING PROCEDURE

Consider a class of recurrent neural networks

given in Fig.1. This network has two layers

with one input and one output. For

convennience, we strictly use mathematical

notations for equations and figures given in

[4]. The input is passed through delays called

TDL. The output is also passed through TDL

and then applied to the first layer. The block

TDL are tapped delay lines. Its output is an

N-dimensional vector, made up of the input

signal at the current time and/or input signal

in the past. IWk,l is an input weight matrixand

LWk,l is a layer weight matrix. Superscripts k

and l are used to identify the source (l)

connection and the destination (k) connection

of layer weight matrices and input weight

matrices.bi, ni, ai, Si and fi are bias vector, net

input, layer output, number of neurons and

transfer function of the layer i (i=1, 2),

respectively. In this case, S2=1 and f2 is a

linear function.

For traditional training, all weights and biases

are updated after each epoch. But for RCNs,

only LW2,l is trained and b2=0. The limitation

of RCNs is that the order of the network is

less than 2. So it can not be applied to identify

systems of higher orders. Thus, we extend its

structure to the network given in Fig. 1. In

addition, based on training method of RCNs

we modify the classical training by fixing

only feedback weights LW1,2 during the

training.


44

Figure 1. A class of recurrent neural networks

By this way, we reduce the number of

training weights. Hence, the training time is

decreased. The initial values for LW1,2 are

chosen in stable areas. The stable domain can

be determined by using stable criteria [5,6,7].

For simplicity, we set these initial values to

small numbers inside the unit circle.

In summary, the proposed training method is

to normally train the given networks using

any training algorithm such as steepest

gradient decent, conjugate gradient and

Levenberg-Marquardt and to keep feedback

weights unchanged during the training. Next,

we apply this training procedure to identify

dynamic nonlinear systems.

SYSTEM IDENTIFICATION

In this section, we apply the new training

procedure to system identifications: a test

problem and a real time system. The first one

is a benchmark problem of a first order

nonlinear system. The latter is a robot arm of

second order system.

For each system, first we create a reference

input R. R is a series of step functions with

random magnitudes and random time

intervals. Then, it is applied to the system.

The output from the system is recorded as T.

Finally, the pair (R,T) is used for

identification training. For amenity, we use

the command trainlm [4] in Matlab to train

the network. Trainlm is written based on the

Levenberg-Marquardt algorithm and back

progapation algorithm [8,9].

The training process includes two stages. The

first stage is called open network training. In

this stage, the feedback line is cut and the

output target T is used as the second input to

the network. Then, the open network has two

inputs and one output. The feedback weights

LW1,2 becomes an input weights IW1,2 and

IW1,2 = LW1,2. The pair ([R;T],T) is used for

open network training. During this training

process, IW1,2 are kept unchanged.

The second stage is called closed network

training or network training. The second input

of the open network is connected to the output

and LW1,2 = IW1,2 are again kept

unchanged.Then, the open network becomes

the closed network or the network. The

trained weights and biases of the open

network is used as the initial values for the

closed network training. First, we divide the

number of samples N into training sets of k

samples (k = 3, 4, ..., N). Next, we train the

closed network with training sets of k

samples. If the mean squared error (MSE) is

small enough (<10-4) then we train the

network with training sets of k+1 samples.

This training process is stopped when k = N-1

or MSE is big.

Test problem

Consider a system given in [10] as follows

3

21

1

y ky k u k

y k

(1)

where u is the input to the system and y is

the output.

For this problem, we use 60 neurons in the

first layer, one delay in the input and one

delay in the feedback output. The number of

weights and biases is 241 while the number of

that in [10] is 460. But the weights, needed to

be trained, are only 181 due to the new

proposed training procedure.


45

The number of samples is N = 100, the same as in [10]. The reference input R is generated

randomly within the interval [-2 2]. The coresponding target T is shown as a blue

curve in Fig. 2. The proposed training procedure is used to train the network. After

training, the network output, the target T and the error are shown in Fig. 2. The error,

difference between the network output and the target, is very small for most samples. In

comparison with the series-parallel identification model [10], there are much

fewer training weights 181 versus 460 and certainly training time is reduced pretty much

because of keeping ¼ of parameters unchanged.

0 10 20 30 40 50 60 70 80 90 100-10

-5

0

5

10

Network Output

Target

0 10 20 30 40 50 60 70 80 90 100-1

-0.5

0

0.5

1

Error

Figure 2. Network output, target and error

Robot arm

In this part, we identify a real robot arm. This

plant is made based on the mathematical model given in [11]. It consists of a link and a

DC motor as shown in Fig. 3. The DC motor is fixed to the end point of the vertical bar. Its

rotor is attached to one end point of the link. The link is made of a white hard plastic bar.

The experimental system includes an Arduino UNO, a dual full-bridge driver L298N, a PC

(laptop) and the robot arm. The Uno is connected to the laptop via USB. A

Matlab/Simulink program is used to generate the reference input to the robot arm through

the Uno and the L298N, and then collect the angle from the encoder attached to the rotor.

Figure 3. A control system for the robot arm

The input is the voltage applied to the DC

motor and the output is the angle of the link

from the vertical bar to the direction of the

link. We generate 2000 random samples of

voltage. Then, this signal is applied to the DC

motor and the angle of the link is measured

via an encoder. The pair of voltage and angle

(radian) signal is used to do identification

training.

The designed network, consists of two layers

with 6 neurons in the first layer. There are one

and two delays in the input and the feedback

output. The total of weights and biases is 37,

but 12 feedback weights are kept unchagned

during the training.

0 200 400 600 800 1000 1200 1400 1600 1800 2000-1

-0.5

0

0.5

1

Network Output

Target

0 200 400 600 800 1000 1200 1400 1600 1800 2000-0.3

-0.2

-0.1

0

0.1

0.2

Error

Figure 4. Network output, target and error


46

After training, the network output, the target T and the error are shown in Fig. 4. The result show that although the error is not small enough for some samples (outside of 0.1 circle), the network output tracks closely the target. This is possibly due to noise, using PWM to change the dc motor’s voltage and low resolution of the encoder (334 pulses/channel).

Conclusions

In this work, a new training procedure is proposed for a class of recurrent neural networks, which is frequently used for identification of dynamic nonliear systems. A benchmark system and a real robot arm are used to test the novel training method. The result shows that we can apply this method to train the class of recurrent neural networks with smaller number of training parameters and a large reduction of training time. Future work focuses on mathematical proof of the proposed method.

REFERENCES

1. Wolfgang Maass, Thomas Natschläger, and Henry Markram,“Real-time computing without stable states: a new framework for neural computation based on perturbations,” Neural Computation, 14(11):2531–2560, 2002. 2. Herbert Jaeger,“The “echo state” approach to analysing and training recurrent neural networks,” Technical Report GMD Report 148, German National Research Center for Information Technology, 2001.

3. M. T. Hagan, H. B. Demuth, and O. D. Jesus, “An introduction to the use of neural networks in control systems,” International journal of robust and nonlinear control, 2002. 4. Howard Demuth, Mark Beale and Martin Hagan, “User’s Guide to Neural Networks Toolbox,” 2013. 5. N. E. Barabanov and D. V. Prokhorov, “Stability analysis of discrete-time recurrent neural networks,” IEEE Trans. on Neural Networks, 2002. 6. M. Liu, “Delayed standard neural network models for control systems,” IEEE Trans. Neural Networks, 2007. 7. Nam H. Nguyen and Martin Hagan, “Stability Analysis of Layered Digital Dynamic Networks Using Dissipativity Theory,” Proceedings of International Joint Conference on Neural Networks, San Jose, California, USA, July 31 - August 5, 2011. 8. Hagan, M. T. and M. Menhaj, “Training feedforward networks with the Marquardt algorithm,” IEEE Transactions on Neural Networks, Vol. 5, No. 6, pp. 989–993, 1994. 9. O. De Jesus and M. T. Hagan, "Backpropagation algorithms for a broad class of dynamic networks," IEEE Trans. Neur. Netw., vol. 18, no. I, pp. 1427, January 2007. 10. Kumpati S. Narendra and Kannan Parthasarathy, ”Identification and Control of Dynamical Systems Using Neural Networks,” IEEE Transactions on Neural Networks, Vol. 1, No. 1, March 1990. 11. Mehmet T. Soylemez, Metin Gokasan and Seta Bogosyan, “Position Control of a Single-Link Robot- Arm Using a Multi-Loop PI Controller,” Proceedings of IEEE Conference on Control Applications, 2003.

TÓM TẮT

MỘT TIẾN TRÌNH ĐÀO TẠO MỚI CHO MỘT PHÂN LỚP

CỦA CÁC MẠNG NEURAL HỒI QUY

Nguyễn Hoài Nam1,*, Trịnh Thị Minh Nguyệt2

1Trường Đại học Kỹ thuật Công nghiệp; 2Trường Cao đẳng Kỹ thuật Yên Bái

Bài báo này đề xuất một thủ tục huấn luyện mới cho một lớp mạng nơ-ron hồi qui. Dựa trên cơ sở

huấn luyện mạng Reservoir, tác giả mở rộng cấu trúc của mạng này từ trễ một nhịp tới trễ nhiều

nhịp và điều chỉnh phương pháp huấn luyện của mạng này. Phương pháp huấn luyện mới được

chứng minh trên một hệ thống chuẩn và một cánh tay máy trong phòng thí nghiệm, và được so

sánh với các phương pháp huấn luyện truyền thống. Kết quả cho thấy, thủ tục huấn luyện được đề

xuất có được một số ưu điểm tốt hơn như: số lượng trọng số và bias được huấn luyện nhỏ hơn và

thời gian huấn luyện có thể nhanh hơn.

Từ khóa: Mạng hồi qui, thủ tục huấn luyện mạng, nhận dạng hệ thống, cánh tay máy một bậc tự do.


Cam Thi Hong Nguyen et al Journal of SCIENCE and TECHNOLOGY 127(13): 47 - 50

47

A NEW STUDY ON OPTIMUM CALCULATION OF PARTIAL

TRANSMISSION RATIOS OF COUPLED PLANETARY GEAR SETS

Cam Thi Hong Nguyen, Trang Van Nguyen, Pi Ngoc Vu*


ABSTRACT This article presents a new study on optimum calculation of the partial ratios of coupled planetary

gear sets for getting minimum radial size of the gear sets. In this article, based on moment

equilibrium condition of a mechanic system including two-row planetary gear sets and their

regular resistance conditions, an efficient model for calculating the partial ratios of coupled

planetary gear sets was proposed. In addition, by giving this explicit model, the partial ratios can

be calculated accurately and simply.

Keywords: Transmission ratio, Gearbox design, Optimum design, Planetary gearbox.

INTRODUCTION*

In gearbox design as well as in planetary

gearbox design, one of the most important

problems is optimum determination of partial

transmission ratios of a gearbox. This is

because the partial ratios are main factors

which affect the size, the dimension, the

mass, and the cost of the gearbox. Therefore,

optimum calculation of the partial ratios of

gearboxes has been subjected to many

studies.

Figure. 1. Schema of a couled planetary gear set

Until now, many researches have been done

on the calculation of the partial ratios of

gearboxes. This type of tasks has been solved

with different gearboxes such as helical

gearboxes (in [1], [2], [3], [4] and [5]), bevel


and bevel – helical gearboxes (in [1], [3], [5]

and [6]) and worm and worm-helical

gearboxes (in [5], [7] and [8]).

Figure. 2. Graph for finding partial ratios

Also, the optimum partial ratios of gearboxes

can be found by different ways: By graph

method (in [1], [3], [8]), by “practical

method” (the ratios were given based on

analyzing practical data (in [5])) and by

models (based theoritical (in [2] and [6]) or

regression models (in [4] and [6]).

From above analysis, it is clear that, there are

many studies have been conducted for

calculating the optimum partial ratios of

gearboxes. However, there was only a study

[1] on this problem for planetary gearboxes.

In the study, the partial ratios of planetary of

three schemas of planetary gearboxes were


48

predicted by graphs. For example, for coupled

planetary gear sets in the Figure 1, the partion

ratio of the low-speed row gear unit pL (pL =

d3L/d1L) can be predicted by the graph in the

Figure 2. After that, the partion ratio of the

high-speed row gear unit pH (pH=d3HL/dH)

can be calculated based on the values of uh

and pL (dw1L, dw1H, dw3L and dw3H are pitch

diameters (mm)).

This article introduces a new study on

optimum calculation of partial transmission

ratios of coupled planetary gear sets for getting

the minimum radial size of the gear sets.

OPTIMUM DETERMINATION OF

PARTIAL TRANSMISSION RATIOS OF

COUPLED PLANETARY GEARBOXES

For the low-speed row of the coupled

planetary gearbox (see Figure 1), the design

equation for the pitting resistance can be

given by the following equation [8]:

1

2

1

2 1(1)

L HL L

HL ML HL L HL

wL w L L

T K uZ Z Z

b d q u

Where, ZML, ZHL and ZL are coefficients

which consider the effects of the gear

material, contact surface shape, and contact

ratio of the first gear unit;T1L is the torque of

the driving shaft (Nm), [HL] is allowable

contact stresses of the low-speed row of the

planetary gearbox.

From (1) the allowable torque (Nm) of the

driving shaft of the low-speed unit can be

found:

22

1

1 2(2)

2 ( 1) ( )

wL w L L L HL

L

L HL ML HL L

b d q uT

u K Z Z Z

From gear theory, we have:

wL baL wLb a (3)

w1L

wL

1

2

Lu da

(4)

Where, bwL and awL are the face with (m) and

the center distance (m) of the low-speed unit.

Substituting (3) and (4) into (2) we get:

3

1 0

14

baL w L L L L

L

d q u KT

(5)

In which:

2

0( . . )

HL

L

HL ML HL L

KK Z Z Z

Put 2

1

( 1)

2

L L

L

L

Z pu

Z

into (5) we get:

3

1 0

1

1

8

baL w L L L L

L

d q p KT

(6)

Calculating in the same way, the allowable

torque of the driving shaft of the high-speed row of the planetary gearbox was found:

3

1 0

1

1

8

baH w H H H H

H

d q p KT

(7)

From (6) and (7), the rate of 1 1/H LT T is:

3

1 0 w1H

1 0 w1L

1

1

H HbaH H H

L baL L L L

T K dq p

T K q d p

(8)

Substituting 1 3 /L w L Ld d p and 1 3 /H w H Hd d p

into (8), the equation becomes:

3 3

1 0 w3H

1 0 w3L

1

1

H HbaH H L H

L baL L L H L

T K dq p p

T K q d p p

(9)

From above equation we have:

33

1

1

1

1

H k H L H

L x L H L

T c c q p p

T c q p p

(10)

In the above equations, 0 0/k H Lc K K ,

3 3/w H w Lc d d , /x baL baHc ; 3w Hd and 3w Ld

are pitch diameters of ring gears of high and low-speed units.

With the coupled planetary gearbox in Figure 1 we have [1]:

1 3

1

1 1H L

H L

T Tp p

(11)

3 1 31 1

H

H L L

H L

pT T T

p p

(12)

From (11) and (12) we get:

1 1

1/H L

H

T Tp

(13)


49

Also, uh can be calculated by the following

equation [1]:

1 1h H Lu u p p (14)

From (14) the the ratio Lp can be given as:

1

1

h

H

L

up

p

(15)

Substituting (13) and (15) into (10) we have: 3 3

2

( 1)( 2)1

( 1) ( 1)

k L L h LH

x L h L

c c p p u pq

c q u p

(16)

In practice, 1 1.2c [1]. In order to get the

minimum radial size of the gearbox 1c [1].

In addition, we can choose 1kc and

/ 1 1.3x baL baHc [8]. Also, the number

of planetary gears of each unit is generally

chosen as 3 3L Hq q . Therefore, equation

(16) can be rewritten as follows: 3

2

( 1)( 2)1

( 1) ( 1)

L L h L

x h L

p p u p

c u p

(17)

To find the value of Lp which depends on the

total ratio of the gearbox hu and the

coefficient xc a computer program was built.

The data used in the program as follows:

15 60hu and / 1 1.3L Hx ba bac . From

the results of the program, the following

regression model (with the coefficient of

determination was 2 0.91R ) was found for

the optimal values of Lp :

0.5141

0.45620.4967 /L HL ba ba hp u (18)

Equation (18) is used to determine the partial

ratio of the low-speed row gear unit of the

gearbox Lp . After finding Lp , the partion ratio

of the high-speed row gear unit Hp

( 3 1/H HH wp d d ) can be determined based on

the values of uh and pL by equation (18) .

CONCLUSIONS

The minimum radial size of the coupled

planetary gearbox can be obtained based on

theoretical analysis and regression method.

Model for calculation of the optimum partial

ratios of doubled planetary gear sets for

getting the minimum radial size of the

gearboxes have been proposed.

The partial ratios of the gearboxes can be

determined accurately and simply by explicit

models.

Acknowledgment

The work described in this paper was

supported by Thai Nguyen University for a

scientific project.

REFERENCES

1. V.N. Kudreavtev; I.A. Gierzaves; E.G.

Glukharev, Design and calculus of gearboxes (in

Russian), Mashinostroenie Publishing, Sankt

Petersburg, 1971.

2. A.N. Petrovski, B.A. Sapiro, N.K. Saphonova,

About optimal problem for multi-step gearboxes

(in Russian),Vestnik Mashinostroenie, No.

10,1987, pp. 13-14.

3. G. Niemann; H. Winter, “Maschinen-

elemente”, Band II, Springer-Verlag ,1989.

4. Romhild I. , Linke H., Gezielte Auslegung

Von Zahnradgetrieben mit minimaler Masse auf

der Basis neuer Berechnungsverfahren,

Konstruktion 44, 1992, pp. 229- 236.

5. G. Milou; G. Dobre; F. Visa; H. Vitila,

Optimal Design of Two Step Gear Units,

regarding the Main Parameters, VDI Berichte No.

1230,1996, pp. 227-244.

6. Vu Ngoc Pi, “A new and effective method for

optimal calculation of total transmission ratio of

two step bevel - helical gearboxes”, International

colloquium in mechanics of solids, fluids,

structures and interactions, Nha Trang, Vietnam ,

2000,pp. 716- 719

7. C.A. Trernapski; G.A. Trekharev, “Design of

Mechanical Transmissions” (in Russian),

Mashinostroenie Publish, Moskova, 1984.

8. Trinh Chat, “Some problems of kinematics

calculation of transmission mechanics system” (in

Vietnamese), Proceedings of the National

Conference on Engineering Mechanics, Vol. 2,

Hanoi, 1993, pp. 7-12.


50

TÓM TẮT NGHIÊN CỨU MỚI VỀ TỐI ƯU HÓA TỶ SỐ TRUYỀN CÁC BỘ TRUYỀN TRONG HỘP HÀNH TINH

Nguyễn Thị Hồng Cẩm, Nguyễn Văn Trang, Vũ Ngọc Pi*


Bài báo này giới thiệu một nghiên cứu mới về phân phối tối ưu tỉ số truyền của các cặp bánh răng

trong truyền động bánh răng hành tinh 2 cấp. Trong bài báo này, dựa trên điều kiện cân bằng mô

men của cơ hệ gồm 2 cấp bánh răng hành tinh và điều kiện sức bền đều của các cấp, các tác giả đã

đề xuất các công thức tính toán tỉ số truyền tối ưu cho từng cấp của hệ. Bằng việc đưa ra các công

thức dưới dạng hàm hiển, tỉ số truyền tối ưu của các cấp có thể xác định một cách nhanh chóng và

đơn giản.

Từ khóa: Tỉ lệ truyền, thiết kế hộp tốc độ, thiết kế tốt ưu, hộp hành tinh.


Thuy Thi Thu Nguyen Journal of SCIENCE and TECHNOLOGY 127(13): 51 - 56

51

ESTABLISHMENT OF A DATABASE OF EMISSION FACTORS FOR ATMOSPHERIC POLLUTANTS FROM STEEL ROLLING PROCESS

Thuy Thi Thu Nguyen*


ABSTRACT

Emission factors have long been used as a cost-effective means to develop area-wide emission

inventories. However, data on emission factors for industries are still scarce in developing

countries including Vietnam. This paper presents results on emission factors of selected air

pollutants being CO, CO2, SO2 and NO for steel rolling process in iron and steel industry.

Concentrations of these pollutants were measured by a combustion gas analyzer, Quintox KM -

9106 at Thai Nguyen Iron and Steel Joint Stock Company. Based on data monitored and the

information of process, total emissions and emission factors of these pollutants for this process

were determined. Computation basing on burning fuel, the emission factors of CO, CO2, SO2 and

NO in the experiment were 0.13 g/kg; 2990 g/kg; 29.36 g/kg and 16.67 g/kg, respectively.

Alternately, base on steel product, the emission factors of CO, CO2, SO2 and NO in the experiment

were 1.89 kg/ton; 42.35 kg/ton; 0.42 kg/ton and 0.24 kg/ton, correspondingly. The results obtained

from the research can be used for industrial emission inventories and air quality management in

Viet Nam.

Keywords: Emission factor, steel rolling, iron and steel, industry, emission inventory.

INTRODUCTION*

An emission factor is a tool that is used to

estimate air pollutant emissions to the

atmosphere. It relates the quantity of

pollutants released from a source to some

activity associated with those emissions.

Emission factors are usually expressed as the

weight of pollutant emitted divided by a unit

weight, volume, distance, or duration of the

activity emitting the pollutant (e.g., pounds of

particulate matter emitted per ton of coal

burned). Emission factors have long been

used as a cost-effective means to develop

area-wide emission inventories. This is a

simple method to estimate pollutant emission

to the atmosphere when the available data are

insufficient [1]. Therefore, establishing of a

database of emission factors for atmospheric

pollutants makes an important contribution to

air quality management in Viet Nam.

Until now, determination of emission factors

in developed countries and international


organizations has significantly developed

(such as AP-42 document of United State, the

database of World Health Organization -

WHO, the document of United Nations

Environment Program – UNEP, the database

of Intergovernmental Panel on Climate

Change - IPCC). Most emission factors of

industries have been applied in these

documents [1,2,21,22]. In Asian countries, the

establishment of emission factors has been

implemented for recent years. However,

determining emission factors in industries has

not yet completed. Most studies on emission

factors focus on coal-fired power industry

[3,4,5]. Generally, the data on emission

factors for iron and steel industry in Asian

countries, including Viet Nam, is scarce and

cannot meet the requirements to control air

pollution which are becoming a necessity for

the countries in the region. The first studies

on emission factors of pollutants in Viet Nam

have been initiated for recent years [6-18].

However, a database of emission factors for

industry in Vietnam is still scarce especially

in iron and steel industry.


52

This study is, therefore, aimed to develop

the emission factors of iron and steel

industry in Vietnamese conditions. This

paper presents preliminary results on the

emission factors of steel rolling for Thai

Nguyen Iron and Steel Joint Stock Company.

MATERIALS AND METHODS

Site description

The study was conducted at the Luu Xa

steel rolling Factory of Thai Nguyen Iron

and Steel Joint Stock Company. After

steelmaking process, steel ingots are heated in

furnace to form into other shapes, such as

blooms, billets, or slabs. Fuel used for baking

is FO oil. The FO oil has ingredient as

follows (C content: 83.4%; H content: 10.0%;

O content: 0.2%; N content: 0.2; S content:

2.9%; ash: 0.3%; moisture: 3.0%).

After passing heat exchanger, the flue gas of

the steel rolling passes through a tunnel

having 70m in length and is directly emitted

into the atmosphere at the height of 60m via a

chimney. This chimney has 6m in diameter at

the bottom and 3m in diameter at the top.

There are two monitoring ports which is

already available to monitor air emission. The

monitoring port no.1 is located at the 1.5m

above the bottom of the chimney and the

monitoring port no.2 is at the following path

from heat exchanger. The monitoring port

no.2 was used in this study.

Monitoring

Monitoring was conducted on November

2013, based on US. EPA Methods 1-3.

Concentrations of selected air pollutants in

the flue gas were directly measured by a

combustion gas analyzer, Quintox KM -

9106 (Kane May, UK). Results are

automatically converted and reported to the

conditions of 1 atm and 0oC by the analyzer.

Detection limits of the equipment for the

monitoring are CO - 0.01%, CO2 –0.01%, SO2

– 1 ppm and NO –1 ppm. Measurement was

conducted for monitoring five times, each

time separated by 20 minutes. Temperature

and velocity of the flue gas were also

measured.

Calculation

+) The concentration of pollutants in the flue

gas: Based on measurement results of

combustion gas analyzer, concentrations of

pollutants were converted to condition 1atm

and 25oC (Viet Nam standard) by the

following the formula:

1 1 1

2 2 2

PV T

PV T (1)

Figure 1. Diagram of the flue gas path and monitoring points of steel rolling process


53

Where: P1, V1, T1 are the pressure, the volume

and the temperature of flue gas at actual

condition, respectively and P2, V2, T2 are the

pressure, the volume and the temperature of

flue gas at standard condition.

+) Emission: The total emission of each

pollutant is calculated by the following

formula:

Total emission = Ci.L.t (mg) (2)

Where, Ci is the concentration of each pollutant

(mg/m3), L is the flow of flue gas (m3/h) and t (h)

is monitoring time.

+) Emission factors: The emission factors of each

pollutant are calculated by the following formula

EFi = Ei

(3) Efuel/Eproduct

Where, EFi is the emission factor of each

pollutant, Ei is the emission of each pollutant

and Efuel/Eproduct are amount of fuel and

product, respectively.

RESULTS AND DISCUSSIONS

+) Emission

Average concentrations of the pollutants in

the flue gas of steel rolling process were

calculated following the formula (1) and

shown in Table 1. As seen from table 1, the

concentration of SO2 exceeded the respective

limit values of QCVN 19: 2009/BTNMT

(column B) [19]. However, CO concentration

in the flue gas was not so high and did

not exceed the respective standard of

QCVN19:2009/BTNMT [19] while the

concentration of CO2 was very high. These

results were very logical to explain. When FO

oil in furnace was sprayed under compressed

air pressure of 2Kg/cm2, the contact area

between oil and hot air would rise and

combustion process took place completely. In

addition, the excess air coefficient which had

greater value also lead to complete

convertibility of the Sulfur and Carbon in the

fuel to SO2, CO2.

The concentration of NO in this study was

consistent with the theoretical concentration

of NOx which approximates 700 mg/Nm3 (600

ppm) when excess air coefficient was 20-25%

[20]. Based on the monitoring data, the total

emission in a shift was calculated by the

formula (2) as shown in Table 2. The time of

this shift was 7.67 hours (460 minutes).

Table 1. Summary information of the flue gas of steel rolling process

Sample Temperature

T (oC)

Flow rate

(Nm3/h)

Concentration

CO

(mg/Nm3)

SO2

(mg/Nm3)

NO

(mg/Nm3)

CO2

(g/Nm3)

CT1 423.8 10970 4.92 1239.84 711.85 129.31

CT2 424.5 10568 5.90 1273.27 718.73 129.31

CT3 422.6 10597 5.90 1272.29 713.82 127.51

CT4 423.3 10194 4.92 1262.46 711.85 125.71

CT5 425.2 10557 6.88 1239.84 712.84 127.51

Average 5.7±0.82 1257.54±6.7 713.8±2.87 127.87±1.50

Table 2. Total emissions of selected of pollutants in steel rolling process

Sample Time

(h)

Flow rate

(Nm3/h)

Emission (g/shift)

CO CO2 SO2 NO

CT1 7.67 10970 413.46 10874969.14 104273.84 59868.57

CT2 7.67 10568 477.98 10476799.73 103164.6 58234.22

CT3 7.67 10597 479.29 10359471.61 103366.14 57993.68

CT4 7.67 10194 384.2 9824832.644 98663.589 55632.74

CT5 7.67 10557 556 10300721.05 100158.8 57585.35

Average 462. 19 10367358.83 101925.39 57862.91


54

Table 3. Emission factors of pollutants in steel rolling process

Sample

Emission factors

CO CO2 SO2 NO

(g/kg Fuel) (kg/ton Steel) (g/kg Fuel) (kg/ton Steel) (g/kg Fuel) (kg/ton Steel) (g/kg Fuel) (kg/ton Steel)

CT1 0.12 1.69 3.13 44.42 30.04 0.43 17.25 0.24

CT2 0.14 1.95 3.02 42.79 29.72 0.42 16.78 0.24

CT3 0.14 1.96 2.98 42.32 29.78 0.42 16.71 0.24

CT4 0.11 1.57 2.83 40.13 28.43 0.40 16.03 0.23

CT5 0.16 2.27 2.97 42.08 28.86 0.41 16.59 0.24

Average 0.13±0.02 1.89±0.27 2.99±0.11 42.35±1.54 29.36±0.69 0.42±0.01 16.67±0.44 0.24±0.01

+) Emission factor

Emission factors of each pollutant were

calculated following the formula (3) and these

results are presented in Table 3.

Data on the emission factors of gaseous

pollutants were compared with other studies

such as the document of UNEP, IPCC. The

emission factor of NO in this study can be

considered to be in the same range with those

of these studies. The emission factor of CO2

in this study was less than a half of the

emission factor published by UNEP (80kg/ton

steel product) [21] while the emission of CO

was twice as high as the one promulgated by

IPCC [22]. The emission of SO2 was ten

times higher than the emission factor

published by IPCC and UNEP but it was

relatively similar to the emission of SO2 of

boiler using FO oil if the fuel has the same

sulfur content. The difference of these results

can be completely understood because

emission factor depends on several factors

including fuel quality, conditions of

combustion/process, air pollution control

system, etc.

CONCLUSIONS

Emission factors of selected air pollutants

being CO, CO2, SO2 and NO for steel

rolling process of Thai Nguyen Iron and Steel

Joint Stock Company were determined.

This is a significant contribution to

Vietnam database of emission factors steel

rolling process in particular and for iron

and steel industry in general. These emission

factors can be used for emissions inventory

and air quality management at the company.

Methodology used in this study can be

applied to other iron and steel industrial

enterprises in the country.

REFERENCES

1. US Environmental Protection Agency

"Compilation of Air Pollutant Emission Factors,

AP-42, 5th Ed", 2002.

2. WHO "Assessment of sources of air, water, and

land pollution, A guide to rapid source inventory

techniques and their use formulating

environmental control strategies"; Part 1: Rapid

inventory techniques in environmental pollution,

1993.

3. Etui-Chan Jean, Someone Myeong, Jae-Whan Sa,

Jinsu Kim, Jae-Hak Jeong, “Greenhouse gas emission

factor development for coal-fired power plants in

Korea”, Applied Energy, Volume 87, Issue 1, January

2010, Pages 205–210.

4. Yu Zhao, Shuxiao Wang, Chris P. Nielsen,

Xinghua Li, Jiming Hao, “Establishment of a

database of emission factors for atmospheric

pollutants from Chinese coal-fired power plants”,

Atmospheric Environment, Volume 44. Issue 12,

April 2010, Pages 1515–1523.

5. S.Nazari, O.Shahhoseini, A.Sohrabi-Kashani,

S.Davari.R.Paydar, Z. Delavar - Moghadam,

“Experimental determination and analysis of CO2,

SO2 and NOx emission factors in Iran’s thermal

power plants”, Energy, Volume 35, Issue 7, July

2010, Pages 2992–2998.

6. Hoang Duong Tung, Nguyen Thi Nguyet Anh

et al, Research on establising emission factors for

pollutant inventory from road transport, Ha Noi,

Environmental Monitoring Centre, 2010

7. Ho Minh Dung, Dinh Xuan Thang,

“Establishment of emission factors from road

transport in Ho Chi Minh city ”, Journal of Labor

Protection, 2009.


55

8. Ta Thu Huong, Le Anh Tuan and Nghiem Trung

Dung, “Preliminary estimation of emission factors for

motorcycles in real-world traffic conditions of Hanoi”,

Journal of Science and Technology, vol. 48, issue 3,

2010, page 101-110

9. Nguyen Thu Trang, Nghiem Trung Dung, Tran

Thu Trang, “Potentiality of co-benefits of climate

and air quality in fuel switching for Hanoi Bus

system”, Journal of Science and Technology, vol. 49,

issue 4, 2011, page 75-86.

10. Nguyen Thi Kim Oanh, Larsbæ Tzreutergårdh

and Nghiem Trung Dung, “Emission of Polycyclic

Aromatic Hydrocarbons and Particulate Matter

from Domestic Combustion of Selected Fuels”,

Environmental Science and Technology, 1999,

pages 2703-2709.

11. Nghiem Trung Dung. Hoang Xuan Co. Pham

Ngoc Ho. Dong Quang Huy. “Emission of

polycyclic aromatic hydrocarbons and particulate

matter from domestic cooking using coal”. Tạp chí

Journal of Applied Chemistry. Vol. 11, 2004. page

31-34

12. Nghiem Trung Dung, Hoang Xuan Co, Pham

Ngoc Ho, Dong Quang Huy, “Emission of

polycyclic aromatic hydrocarbons and particulate

matter from domestic cooking using sawdust”,

Journal of Science and Technology, vol. 43, issue 1,

2005, page 108-113

13. Nghiem Trung Dung, Nguyen Viet Thang,

“Determination of emission factors for domestic

sources using biomass fuel”, Journal of Science

and Technology, No 82A, 2011, pages 32-36

14. Nghiem Trung Dung, Le Phuong Thuy,

“Determination of emission factors for domestic

sources using fossil fuel”, Journal of Science and

Technology, No 82A, 2011, pages 32-36

15. Nghiem Trung Dung. Larsbæ Tzreutergårdh.

Nguyen Thi Kim Oanh. Dang Lim Chi. Hoang

Xuan Co. “Emission of polycyclic Aromatic

Hydrocarbons associated with particulate matter

from a coal-fire power plant in Vietnam”. Journal

of Applied Chemistry. Vol. 9, 2004. page 36-40

16. Nguyen Thi Thu Thuy “Determination of

emission factors for steel rolling using FO oil”,

Thai Nguyen Journal of Science and Technology,

No 12, 2011

17. Nguyen Thi Thu Thuy, Nghiem Trung Dung,

“Determination of emission factors for coke

quenching”. Journal of Science and Technology,

No 87, 2012, pages 62-66

18. Nguyen Thi Thu Thuy, “Determination of

emission factors for coke quenching and sintering

in iron and steel industry”, International

Symposium on Technology for Sustainability

Workshop, Bangkok, Thailand, 2012.

19. MONRE. “QCVN 19:2009/BTNMT - National

Technical Regulation on Industrial Emission of

Inorganic Substances and Dusts”, Hanoi, 2009.

20. Noel de Nevers, "Air pollution control

engineering", McGraw-Hill, 1995

21. Department of Environmental impact

assessment and appraisal - Vietnam Environment

Administration, “The guide to preparing

Environmental impact assessment reports in cast-

iron and steel plants”, Hanoi, 2009 [22] IPCC

www.ipcc-nggip.iges.or.jp, access date 20/06/2014.


56

TÓM TẮT

XÂY DỰNG DỮ LIỆU BỘ HỆ SỐ PHÁT THẢI

CÁC CHẤT Ô NHIỄM KHÔNG KHÍ TRONG QUÁ TRÌNH CÁN THÉP

Nguyễn Thị Thu Thủy*


Từ lâu hệ số phát thải đã được coi như một công cụ hiệu quả về kinh tế trong kiểm kê khí thải. Tuy

nhiên, dữ liệu về hệ số phát thải còn khá khan hiếm tại các nước đang phát triển trong đó có Việt

Nam. Nghiên cứu này được thực hiện để xác định hệ số phát thải của CO, CO2, SO2 và NO trong

quá trình cán thép. Nồng độ các chất ô nhiễm được đo đạc bằng thiết bị đo nhanh - Quintox KM-

9106 - tại công ty Gang thép Thái Nguyên. Tổng lượng phát thải chất ô nhiễm và hệ số phát thải

của từng chất đã được tính toán dựa trên kết quả quan trắc và thông tin về quá trình. Tính theo

nhiên liệu, hệ số phát thải của CO, CO2, SO2 và NO tương ứng là 0,13 g/kg; 2990 g/kg; 29,36 g/kg

và 16,67 g/kg. Theo sản phẩm, hệ số phát thải của CO, CO2, SO2 và NO tương ứng là 1,89 kg/ton;

42,35 kg/tấn; 0,42 kg/tấn và 0,24 kg/tấn. Kết quả đạt được trong nghiên cứu này có thể được sử

dụng trong kiểm kê phát thải công nghiệp và quản lý môi trường không khí tại Việt Nam

Từ khóa: Hệ số phát thải, cán thép, dầu FO, ô nhiễm không khí, kiểm kê phát thải.


Duy Nguyen Tien et al Journal of SCIENCE and TECHNOLOGY 127(13): 57 - 62

57

DISSOLVED OXYGEN CONTROL OF THE ACTIVATED SLUDGE WASTEWATER TREATMENT PROCESS USING HEDGE ALGEBRAICCONTROL

Duy Nguyen Tien, Binh Lam Hoang, Son Lam Hung, Huy Nguyen Phuong*


ABSTRACT

Activated sludge wastewater treatment process is difficult to be controlled because of its complex,

time-varying and non-linear behavior. In this process, the control of the dissolved oxygen (DO)

concentration in the reactors plays an important role in the operation of the facility.In addition to

evaluating some controllers for DO such as PI, FLC, this paper also introduces a new approach,

using Hedge Algebraic (HA). The results show that Hedge Algebraic Controller (HAC) can be

effectively used for DO control in activated sludge wastewater treatment process.

Keywords: Hedge Algebraic control, Dissolved Oxygen control, Wastewater treatment.

INTRODUCTION*

Aeration is an important part of the whole

activated sludge wastewater treatment process

because aerobic conditions are conducive to

the growth of a wide variety of microbes,

including heterotrophic bacteria, which

remove biochemical oxygen demand (BOD)

from the wastewater, as well as nitrifying

bacteria, which oxidize ammonia to nitrate.

Because of the strong and fast effects of

aeration on biomass growth, DO control is the

most studied control problem in wastewater

treatment 1. DO control is the difficulty and

focus of biological wastewater treatment

processes. Insufficient or excess oxygen in

the aeration tank will lead to the deterioration

of activated sludge.

In the past, most of the DO control systems

reported in literature employ classical PI

controllers, which are mainly because of the

inherent simple design and application, and

mostly satisfactory performance in the

operating range of interest.

Fuzzy controllers (FLC) have much superior

in quality compared with PI controller.It is

often pointed out that FLCs do not require

knowledge of a detailed mathematical model

of the control system and allow a more

intuitive approach to design compared to the


PI controller because the fuzzy sets used in

fuzzy control aim to capture the semantics of

natural linguistic terms present in the fuzzy

controller knowledge. In addition, they have

the capability of handling uncertain and noisy

signals, and usually lead to better results

compared to the classical PI controller in

terms of response time, settling time and

robustness 5.

Hedge algebras were developed in 2 to model

the order-based semantics of the terms in

term-domains of linguistic variables. Then,

the fuzzy rules can be viewed as to define

points in a Cartesian product of suitable

hedge algebras, and approximate reasoning

method on the controller knowledge can be

transformed into an interpolation method on a

real surface defined by these points by using

fuzziness parameters values 3. Since this

transformation is defined by Semantically

Quantifying Mappings (SQMs) of hedge

algebras, which may preserve the relations

between the variables based on the order-

based semantic of terms in the controller

knowledge, the resulting surface can be

considered as an appropriate mathematical

model of the controller knowledge. So, hedge

algebras may provide a sound formalized

basis to develop effective new reasoning

methods for a kind of controllers, called

hedge algebra (HA) controllers.


58

In this paper, in addition to evaluating some

controllers for DO such as PI, FLC, this paper

also introduces a new approach, using HAC.

The results show that HAC cang be

effectively used for DO control in activated

sludge wastewater treatment process.

HEDGE ALGEBRA CONTROLLER OVERVIEW

Hedge algebras (HAs) are aimed to show that

the inherent ordered-based structures of term-

domains of linguistic variables are useful to

discover order-based semantic properties of

terms and term-domains 23. On this

viewpoint, every term-domain of a linguistic

variable X can be considered as an HA, AX =

(X, G, C, H, ), where X is a term-set of X;

is an order relation on X, which is regarded as

to be induced by the inherent order-based

semantics of the terms of X; G = {c–, c+},

where c– (or, c+) is called the negative (or,

positive) primary term, is the set of generators

that satisfy c– c+; C = {0, W, 1} is the set of

constants satisfying 0 ≤ c– ≤ W ≤ c+ ≤ 1,

whose meanings state that 0 and 1 are,

respectively, the least and the greatest term in

X, W is the neutral term; H = H– H+, where

H– = {hj: 1 j q} is the set of negative

hedges hH satisfying hc+ c+ (written as

sign (h) = 1) and H+ = {hj: 1 j p} is the

set of positive hedges h satisfying hc+ ≥ c+

(written as sign (h) = +1). Since hj’s are

regarded as unary operations, every term of

AX, except from the constants, is of the form

hnhn-1 ... h1c, c G. Many inherent semantics

properties of terms and, especially, hedges

can be discovered in the structure of AX. For

instance, hx and x are always comparable, for

every x X and h H; assuming that hx ≥ x,

the comparability of hx and khx implies that

either khx ≥ hx ≥ x, which is indicated by

sign(k, h) = +1, or hx ≥ khx ≥ x, which is

indicated by sign(k, h) = 1. E.g. we can

check that sign(V,L) = +1, as Vlbig ≤ Lbig

≤big, while sign(V,R) = 1, as

Rbig≤VRbig≤big. Then, every xX\C has a

sign defined by Sgn(x) = sign(hn, hn-1) …

sign(h2h1)sign(h1)sign(c), where x = hnhn-1 …

h1c, for c G. It is proved that Sgn(hx) = 1

hx ≤ x and Sgn(hx) = +1 hx ≥ x.

The semantic structure of AX discovered in

the algebraic approach to the term semantics

implies that the set HI(x) = {x = hnhn-1 ... h1c :

cG, hjH }{x}, for every xX, can be

considered as the model of the fuzziness of x.

The structure of the set of all such sets, H(x),

xX, induces a fuzziness measure fm of the

terms of X, which is equal to the “diameter”

of H(x) and can be calculated by given

fuzziness measure of the primary terms,

fm(c) and fm(c+), and the fuzziness measure

of hedges, (h), h H, called commonly the

fuzziness parameters of X. We have that for

every x X,x = hnhn-1 ... h1c,

Gccfmhhchhhfm nnn ),()()...()...( 111 (1.1)

In turn, a given fuzziness measure fm of X

induces numeric term semantics, defined by

the so-called Semantically Quantifying

Mapping (SQM) fm, which is also calculated

by the given fuzziness parameter values as

follows:

( )

( ) ( ), ( ) ( )

( ), ( ) ( )

( ) ( )

( ) ( ) ( ) ( )

fm fm

fm

fm j fm

j

j i j ji sign j

W fm c c fm c

fm c c fm c

h x x

Sgn h x fm h x h x fm h x

(1.2)

where

,))(()(12

1)( xhhSgnxhSgnxh jpjj

for all integers j[q^p]=[q, p]\{0}.

DESIGN HAC FOR DISSOLVED OXYGEN

CONTROL

Dissolved Oxygen Control of the Activated

Sludge Wastewater Treatment Process

The general overview of activated sludge

wastewater treatment process is shown in

Fig.1. The inflow is first processed in the

bioreactor where, by the action of

microorganisms, the substrate content is

reduced. Next, the water flows to a settler,

where the biomass sludge is recovered. The

clean water remains at the top of the settler

and is carried out of the plant, and a fraction

of the sludge is returned to the input of the

bioreactor in order to maintain an appropriate

level of biomass, allowing the reduction of


59

the organic matter. The rest of the sludge is

purged 9.

A block diagram of the DO controller is

shown in Fig2. According to [8], DO transfer

function is given by:

(2.1)

Where K=0.8, T1=12, T2=100, =60.

In order to verify responses of these

controllers when changing of DO, the

expected value of the DO areset in Table 1.

Determination of hedge algebras for the

linguistic L-variables and linguistic rule base

First, it is necessary to specify the same sets

G, C and H for all the required HAs as

follows:

1) The set G of the primary terms is assumed

to be {S, B}, i.e. the negative primary term

c = S (Small) and the positive primary term

c+ = B (Big).

To express the semantics of terms present in

fuzzy model, it is sufficient to define

H– = {L (Little)} and H+ = {V (Very)}.

The whole term-set X of every HA is then

completely determined, regardless its term-set

is finite or infinite. Examples of terms are B

(Big), VB (Very Big), S (Small) and so on.

Then, the HA-terms with length of one or two

is established as in Table 2. However, since

there are only five labels of the variables Le

and Lce, their term-transformations are

established. All the established transformations

should preserve the order-based relationships

and the opposite meaning of terms, e.g. the

opposite terms VS and VB are of opposite

meaning.

Table1. The expected value of the DO with time

t [s] 0 – 200 200– 400 400 – 600 600 – 800 800 – 1000

*

dcu [mg/l] 2 3 4 1 2

Figure 1. General overview of activated sludge wastewater treatment process8

Figure 2. A block diagram of the DO controller

Controller Air-blower

Oxygen mass

transfer

de/dt

+ -

ec

u y


60

Table 2. Term transformation of the linguistic

variable Lu

For FL NB N ZE P PB

Linguistic

Variable Lu VS S W B VB

linguistic

variables Le

and Lce

S LS W LB B

Determination of the mathematical model

of the given rule base for desired HA

controller

The HA rule base is represented in Table 3.

The linguistic values of their cells can be

determined based on the monotonicity of the

linguistic terms in either the respective

column or row, i.e. they can be drawn

intuitively from the given rule base.

Table 3. Rule base for the HA controller

Lce

Le S LS W LB B

S VS VS VS S W

LS VS VS S W B

W VS S W B VB

LB S W B VB VB

B W B VB VB VB

To determine the mathematical model of Sreal

of the rule base with HA-terms given in Table

3, the fm-values of the terms of every

variable present in Table 4 must be computed

by applying Eqs. (1.1) and (1.2), examined in

more detail in 5, for determining SQM fm,

induced by a given fuzziness measure fm of

every variable. However, here for the terms in

Table 3, which are of only length not greater

than 2 and generated by the use of only two

hedges in H = {h-1, h1}, i.e. every term

contains at most one hedge which is either h-1

or h1, Eqs. (1.1) and (1.2) can be rewritten

simply as follows, noting that, as i = sign(j) =

j {-1, 1}.

Calculated with Eq (1.1) and Eq (1.2), a grid

of the points of the surface Sreal corresponding

to all the linguistic points defined by the cells

of Table 4 can be calculated and obtained as

presented in Table 5. Now, an ordinary

interpolation method on Sreal can simply be

selected as an HA-IRMd. It can be observed

observe again that once an HA-IRMd is

selected, all components of the designed HA

controller can completely be determined by

merely the independent fuzziness parameters

of the variables Le, Lce and Lu. Thus, as there

are not many numeric interpolation methods

to be selected, it is useful that to enhance the

performance of the designed controller it is

only necessary to determine appropriate

fuzziness parameters values of the linguistic

variables, as presented in Table 4.

Determination of the quantitative semantics

of terms for each linguistic variable

Once the HAs of the variables are determined,

the determination of their quantitative term

semantics is simply to specify values of the

fuzziness parameters of the variables. For

simplification, these parameter values were

determined in Table 4, noting that always

fm(B) + fm(S) = 1 and μ(L) + μ(V) = 1. Using

these parameters values all the terms in Table

4 can be computed, which means that the

surface Sreal is determined approximately.

Table 4. Determined parameters Values of Le,

Lce and Lu

Le Lce Lu

fm(S) 0.5 0.5 0.5

= μ(L) 0.3 0.3 0.7

All the linguistic points represented in Table

3 are transformed into the points shown in

table 5.

The points shown in Table 5 determine

approximately the surface Sreal as represented

in Fig. 3.

Figure 3. Surface Sreal for the HAC


61

Table 5. SQM-values of the HA-terms

e

ce

0.350

(S)

0.455

(LS)

0.500

(W)

0.545

(LB)

0.650

(B)

0.350

(S)

0.045

(VS)

0.045

(VS)

0.045

(VS)

0.150

(S)

0.500

(W)

0.455

(LS)

0.045

(VS)

0.045

(VS)

0.150

(S)

0.50

(W)

0.850

(B)

0.500

(W)

0.450

(VS)

0.150

(S)

0.500(

W)

0.850

(B)

0.955

(VB)

0.545

(LB)

0.150

(S)

0.500

(W)

0.850

(B)

0.955

(VB)

0.955

(VB)

0.650

(B)

0.500

(W)

0.850

(B)

0.955

(VB)

0.955

(VB)

0.955

(VB)

Simulations

In order to evaluating effectiveness of HAC for

DO, we also design PI control and FLC.

Parameters of the PI controller for DO is

KP = 5.2, TI = 0.1. Parameters of the FLC is

corresponded to HAC. The results are showed

in Fig 4.

Fig 4 shows responses of these controller to

the change of the reference DO over time. For

the classic PI controller, the response has the

largest overshoot/undershoot and the longest

establishment time. These values were

significantly reduced for FLC. And

especially, the HAC response is the best of

overshoot/undershoot and establishment time.

CONCLUSIONS

In this paper, a new type of controller based

on hedge algebra was proposed and applied

for Dissolved Oxygen Control of the

Activated Sludge Wastewater Treatment

Process. To our best knowledge, this is the

first time that the hedge algebra principles

were successfully applied in this field. The

effectiveness of the proposed approach was

evaluated through comparison with both the

classical PI controller and the FL controller.

The simulation was performed and the results

were gathered in reasonably wide ranges of

the DO. Based on the obtained results, the

developed HA controller proved to be the best

choice for the purpose out of the considered

controllers because of its superior

performance in terms of tracking the

reference DO and robustness. More

specifically, it was shown that only the HA

controller manages to reconcile the

requirements for fast response and small

overshoot/undershoot. At the same time, its

computational efficiency was shown to be

satisfactory and similar to that of the other

two considered controllers. The newly

developed HA controller was definitely

proven to be research worthy and a promising

alternative to the existing solutions in the

field of control engineering.

Figure 4. Simulation results of PI controller, FLC, HAC for DO


62

REFERENCE

1. C. A. C. Belchior, R. A. M. Araújoa, J. A. C.

Landeckb, Dissolved oxygen control of the

activated sludge wastewater treatment process

using stable adaptive fuzzy control, Computers

and Chemical Engineering, 37: 152-162, 2012.

2. C.H. Nguyen, W. Wechler, Hedge algebras:

An algebraic approach to structures of sets of

linguistic domains of linguistic truth variable,

Fuzzy Set. Syst., 35 (1990) 281-293.

3. C.H. Nguyen, N. L. Vu, X. V. Le, Optimal

hedge-algebra-based controller: Desgn and

application, Fuzzy Set. Syst., 159 (2008) 968-989.

4. C.H. Nguyen, N. L. Vu, X. V. Le, Quantifying

hedge algebras, interpolative reasoning method

and its application to some problems of fuzzy

control, WSEAS T. Comput., 5 (2006) 2519-2529.

5. C.H. Nguyen, D.A. Nguyen, N.L. Vu, Fuzzy

Controllers Using Hedge Algebra Based

Semantics of Vague Linguistic Terms, in: D.

Vukadinović (Ed.), Fuzzy Control Systems, Nova

Science Publishers, Hauppauge, 2013, pp. 135-

192.

6. J.M. Zurada, R.J. Marks, C.J. Robinson,

Computational Intelligence Imitating Life, IEEE

Press, Piscataway, 1994.

7. S.R. Buss, 3D Computer Graphics: A

Mathematical Introduction with OpenGL,

Cambridge University Press, New York, 2003.

8. YE Hong-tao, LI Zhen-qiang, LUO Wen-

guang, Dissolved Oxygen Control of the Activated

Sludge Wastewater Treatment Process Using

Adaptive Fuzzy PID Control, Proceedings of the

32nd Chinese Control Conference, pp 7510-7513,

2013

9. W. C. Chen, N. B. Chang, J. C. Chen, Rough

set-based fuzzy neural controller design for

industrial wastewater treatment, Water Research,

37 (1): 78-90, 2003.

10. Jain, L.C., and Jain R.K.: “Hybrid

Intelligent Engineering Systems”, World

Scientific Publishing, 1997.

TÓM TẮT

ỨNG DỤNG ĐẠI SỐ GIA TỬ ĐỂ THIẾT KẾ BỘ ĐIỀU KHIỂN

LƯỢNG OXY HÒA TAN TRONG QUÁ TRÌNH XỬ LÝ NƯỚC THẢI

THEO PHƯƠNG PHÁP BÙN HOẠT TÍNH

Nguyễn Tiến Duy, Lâm Hoàng Bình, Lâm Hùng Sơn, Nguyễn Phương Huy*


Xử lý nước thải bằng phương pháp bùn hoạt tính là một quá trình điều khiển khó khăn do hệ thống

phức tạp, biến đổi theo thời gian và có tính chất phi tuyến. Trong quá trình này, việc điều khiển

lượng oxy hòa tan (DO) tập trung ở các lò phản ứng đóng vai trò rất quan trọng trong hoạt động

thực tế của hệ thống. Trong bài báo này, ngoài việc đánh giá việc điều khiển DO theo các phương

pháp như PI, FLC, các tác giảcòn giới thiệu một cách tiếp cận mới, sử dụng đại số gia tử (HA).

Các kết quả thu được cho thấy các bộ điều khiển sử dụng đại số gia tử kiểm soát hiệu quả hơn

nồng độ DO trong quá trình xử lý nước thải theo phương pháp bùn hoạt tính.

Từ khóa: Bộ điều khiển dùng đại số gia tử, điều khiển oxy hòa tan, xử lý nước thải.


Khuyen Thi Minh Pham et al Journal of SCIENCE and TECHNOLOGY 127(13): 63 - 68

63

SUPPLY CHAIN MANAGEMENT FOR COLLEGES/UNIVERSITIES: SOLUTIONS TO IMPROVE THE EFFICIENCY OF THE TRANSFER OF SCIENCE AND TECHNOLOGY

Khuyen Thi Minh Pham*, Yen Thi Mai Pham


ABSTRACT

The Integrated Educational Supply Chain Management (IESCM) for the colleges and universities

provides two main contributions to the society, including human resource contribution and

research contribution through two main activities of colleges/universities: education and research.

In the world, theories about Educational Supply Chain were studied in the last years of the 90s,

but in Vietnam these contents are still new. This article depicts a holistic view, comprising inputs,

the process, and outputs of the educational supply chain. With the analysis about three group

activities of supply chain management and some experiences from education developing

countries, the paper also provides educational management a new dimension to understand how

supply chain management contributes to successful university operations, especially enhances the

efficiency of scientific and technological transfer.

Keywords: The Integrated Educational Supply Chain Management (IESCM), Efficiency of scientific and

technological transfer, Relationship Management, Social Orientation, Strategy.

INTRODUCTION

In recent times, the theory of supply chain

management (SCM) has been widely studied

under a variety of labels and for a number of

reasons: improving operations, better

outsourcing, increasing profits, enhancing

customer satisfaction, generating quality

outcomes, tackling competitive pressures,

increasing globalization, increasing

importance of E-commerce, and growing

complexity of supply chains.

In the world

The first theory about supply chain

management appeared in 1950s with the

development of goods and relationships in

businesses... and it rapidly widened their

applications to other fields of life. (Fig 1)

It is a surprising fact that researchers

developed SCM models focusing mostly on

improving business operations. Few,

particularly academic researchers, do not

realize that the research on academic supply

chain management may also be conducted for

their own educational institutions.

Figure 1. Timeline of theory in Supply chain management*

* Tel: 01688140486Email: [email protected]

195

0

197

0

199

0

201

0

Initiated Logistic Concept Initiated the SCM Concept

1980

Matured Logistic Concept 1985

SCM in the Manufacturing Industry

1995

Initiated SCM in the Service Industry

2007

Education SCM

ITE SCM

2009

Redesigned ITESCM

2012


64

Until 1996, O’Brien [1] proposed an

educational supply chain as a tool for

strategic planning in tertiary education. The

study was based on a survey among

employers and students. Survey findings

revealed that integration and coordination

among students and employers should have

been promoted. In 2007, Lau [2] performed

an in-depth case study approach to developing

an educational supply chain management for

the City University of Hong Kong. In fact,

Educational supply chain represents

international supply chain management

concept as it is the uniform dimension for the

international arena (Habib, 2008-2012) [3].

In Vietnam

The concepts of SCM are rather new with

Vietnamese business. The Manager started

considering the theory of SCM in some recent

years when Vietnam became a member of

WTO and they had to face with the strong

competition of global companies. But, almost

researches focused on the SCM in

manufacturing industry.

LITERATURE REVIEW

Model of supply chain for colleges/

universities

The objective of the educational supply chain

is to develop the quality graduates and

researches with limited resources for the

society, which is the final customer or

consumer. To achieve this goal, educational

institutions need to have a certain degree of

knowledge about the partners in their supply

chains including suppliers, customers, and the

consumer. (Fig 2)

Suppliers

Education Suppliers include: Suppliers of the

student (high schools/colleges); Suppliers of

the faculty (Other universities/colleges);

Source of fund – Family (Parents, siblings),

relatives, etc; Government and private

organizations (Scholarship); Suppliers of

assets or equipment (Furniture, computer,

networking equipment, etc.); Suppliers of

educational materials (Stationery, instruction

materials, etc.);

Research Suppliers include: Suppliers of

Internal Research Projects (University self

funding); External research projects (External

research funds, Ministry of Education, private

organizations, etc.)

Figure 2. Model of Educational Supply chain for Colleges/universities [4]

Graduates Research

Education customers Research Customers

Society

Education Supplier Research Supplier

Pupils Research

projects

Academic Strategies, plans,

operation and quality

Research Strategies, plans,

operation and performance

evaluation.

Colleges/Universities

PE UC FC FA PE UC FC FA PE: Programs Establishment

UC: Universities culture

FC: Faculty Capabilities

FA: Facilities


65

Activities of colleges/universities

A university is regarded as a service provider

with 2 main services: Education and

Research. Through proper educational

management, the university can produce

quality outcomes for the society.

Factors influence on the supply chain for

colleges/universities often are Programs

Establishment (PE), Universities culture

(UC), Faculty Capabilities (FC), Facilities

(FA). In Vietnam, it also be depend on the

management and regulation of government.

Outcomes

Graduates: Quality of graduates often

include: Knowledge (Tacit or Explicit), skills,

competencies, capabilities, ethics, career

Development Programs

Researches: Quality research outcomes may

include problem solution, pure theory, thesis

findings, internal and external projects

applications, researchers, research

publications, or research findings, etc.

Customers

EEducation Customers: Graduates, family

(parents, siblings, relatives, etc.),

employers of government and private

organizations.

Research Customers: Funding organizations

of research projects, research outcomes

(researchers, research publications, findings

etc.), Others (research professional

organizations, Society of manufacturing

engineers and Trade Associations, etc.).

Consumers

The society is the end customer in

educational supply chain. As

colleges/universities are the part of the

society, the final outcomes of this supply

chain, including graduates with desirable

quality and quality research outcomes are

delivered to the society.

The activities in supply chain management

for colleges/universities and their effects to

the transfer of science & technology

There are three groups of activities in supply

chain management for Colleges/universities.

Operate the activities of Colleges/universities with Society orientation

In educational management, three decision levels are involved in the process of the university: 1. Strategic Level- general direction, long-term goals, philosophies and values; 2: Planning Level: decisions support strategic decisions; 3: Operating Level: every day decisions, Operational decisions can be pre-programmed, pre-made, or set out clearly in policy manuals. These decision should be based on the social need for human resources and researches and abilities of colleges/universities.

Relationship management with suppliers

Raw material in educational supply chain for colleges/universities are pupils in high school and other colleges and they take an importance part in the success of education process. Pupils need to be career oriented early and clearly to have good preparation and attitude to study in colleges/universities . This is only achieved when universities and schools have a good relationship with each others. Good relationship with high schools also help colleges/universities have advantage in attracting pupils.

Research Suppliers are specially important with research supply chain for colleges/universities because they supply fund and information for research. colleges/universities want to become research colleges/universities need have a big self research funding and lots of External research funds.

Relationship management with customers

Relationship management between colleges/universities and Businesses represented by their co-operations:

1. Cooperation in Research: The purpose of this collaboration is to achieve support for research activities of the colleges/universities, implementing projects that link the academic world and the business conduction.

2. Commercialization of research results: It includes technology transfer. In developing countries like Vietnam, to be able to promote this form of cooperation, it is necessary to do immediately is to strengthen the institutional framework to ensure the actual intellectual property rights.


66

3. Promote mobility of students: by creating mechanisms to support them, such as putting students in internships and creating opportunities for them to be able to experience many rich aspects of the world outside the colleges/universities . Strengthen coordination with the human resource department of the company, business to facilitate students to the world of work.

4. Promote the movement of academia: Encourage exchanges or short-term contract work of academics in the business to build relationships, share their views and capture reality.

5. Develop and implement training programs: Improve the quality of education and help students adapt well to the demands of the labor market. It should encourage the involvement of businesses in the construction and renovation of colleges/universities programs, through discussion and information exchange.

6. Support business start-up activities: creating a stimulating cultural faculty and students to think and act with entrepreneurship, put them to the way of the corporate world and attract them to escape the old path of thinking.

These co-operations between businesses and colleges/universities have contributed to social many benefits: Creates new knowledge (through research, reflected by patents, inventions, scientific articles), promotes production (through technology transfer, expressed through economic growth based on knowledge), providing skilled human resources (through training, demonstrated by the number of graduates have jobs).

EXPERIENCES FROM OTHER COUNTRIES

The United State (U.S): High investment from Government and Companies to colleges/universities; Good relationship between high schools, colleges/universities and business

In the U.S., the U.S. government's investment for Science and Technology higher than the same investment in the governments of European countries and Japan combined, while investment in Science and Technology

of the U.S. company 3 times higher than the value of government investment. The policy of the U.S. government is creating conditions for all U.S. citizens to access to education and training they need.

U.S. universities have good relationship

between high schools and Business and its

result is that, currently, one third of the research

in the world of Science and Technology

published annually branded America. 45% of

professors, scientists teaching at U.S.

universities are foreign nationals.

Singapore: colleges/universities are real

businesses, always orient to social [5]

Singapore is well-known as a center of high

technology, have contributed of the

university at international level. They handle

very well the relationship between

universities with businesses around the

exploitation of research results and protection

of intellectual property rights. This experience

is significantly useful for Vietnam. Being the

country's only two full-fledged universities up

to the turn of the century, the National

University of Singapore and Nanyang

Technological University have a strong

tradition of collaborating with industry. Their

graduates continue to find employment

readily in the diverse manufacturing and

service sectors in the country. The culture of

interaction with industry has been developed

through a range of activities including

internships, research collaboration,

technology licensing, adjunct appointments

and industry participation in consultative

committees of academic departments.

LESSONS FOR VIETNAM’S

COLLEGES/UNIVERSITIES

To improve the efficiency of scientific and

technological transfer we need to care about

two main issues: 1) The relevance of science

and technology is transferred to reality; 2)

The capacity of the parties to perform the

transfer. The training institutes in the world

solve these problems by managing the

relationship between the members of the

educational supply chain. In Vietnam, it is

still new lessons.


67

Improve quality of education and research with social orientation

Concentrate on Students is outdated. Enhance education and research quality is not only about improving level of lectures, education programs and services but also all of them to satisfy the social needs.

Many managers of colleges/universities still think that they need to concentrate on needs of students such as their knowledge, their accommodation, their leisure activities,… But, the main and most important purpose of learning in colleges/universities is to get a good job for one’s life. And we can do it only by giving them the knowledge, skills and attitude that the society needs. The number of graduates that have jobs in properly trained field is an important criterion that made the reputation of a institution

The reputation of one College/University also depends on their abilities on research. colleges/universities in Vietnam. The universities of Vietnam are often underestimated in research capacity. We have a lot of professors, doctors but the numbers of researches which are recognized in the region and the world are even lower than the average level of the Southeast Asia region. There are two main reasons are: 1) Lack of orientation mechanisms, support scientific research at both national and Universities levels. 2) Separation to research environment (social demand) of scientists. So, they are the urgent problems that need to change in order to improve the efficiency of scientific and technological research.

Relationship between Colleges/universities and high school

Study the social needs, design right product, show what we need from the supplier and help them supply good raw material is the process to make a suit product and gain a better benefit in a Supply chain. And this process also apply in Educational Supply chain. Help high schools orient the career for the pupils basing on their abilities and dream work to chose right pupils with their love and abilities to study will help colleges/universities have best raw material for education. These graduates will be good

employee for scientific and technological transfer process from colleges/universities to society with good abilities, knowledge and attitude.

We can do it by exchanging experience in developing training programs, practicing View and experience programs in colleges/universities for pupils and promoting the reputation and activities of colleges/universities.

Relationship between Colleges/universities and business/organizations

Stay together to talk and share the view to cooperate

The relation will bring many advantages for the partner: The Universities have graduates working more efficiently and better usability scientific research; Enterprises have more quality human resources to carry out the process of applying science and technology to be transferred from the theoretical model.

But, in Vietnam, Colleges/universities and businesses have little connection with each other. The Colleges/universities just try to expand the size and training programs from the model of famous universities in the world and seek to attract students. However, very few of them assess training programs that if they actually fit with the requirements of business and organizations that recruit their graduates. There is very little real concern of universities to make the relationship with business and society to encourage scientific research in accordance with practical requirements. The business always criticize the universities for their theoretical and ineffective program but they themselves accept retraining costs rather than invest in supporting for training and research activities. So universities and businesses need to sit back and interact more on what they need to create real benefits for both parties. They need to build an effective partnership mechanisms including the terms of the rights and responsibilities of each party.

Organize the centre to manage the relationship

The center will organize and implement activities to manage the Colleges/ Universities’ relationship. The main activities of the centre include:


68

1) Construct the Funds for research from Colleges/universities and Businesses;

2) Cooperate in research and transfer of science and technology: Take place programs to exchange experience and knowledge of science and technology; Develop mechanisms for intellectual property rights of the products of science and technology.

3) Organize the study and work programs for students and faculty in business: lessons from lectures of Business, visiting and practicing programs in business .

4) Cooperate in evaluating the effectiveness of training programs and research activities.

CONCLUSION

From the article, the interrelationships among all educational management components are investigated. It also show that the Integrated Educational Supply Chain Management will give Colleges and Universities a chance to enhances the efficiency of scientific and technological transfer in theory and experiences from others countries. With Colleges and Universities in Vietnam, they need to:

1) Improve quality of education and research with social orientation.

2) Manage the relationship with high school in job orientation and building the programs.

3) Manage the relationship with business/organizations in education and research.

REFERENCE

1. O’Brien, Elaine M. and Kenneth R. (1996),

“Educational supply chain: a tool for strategic

planning in tertiary education?” Marketing

Intelligence & Planning, Vol. 14 No. 2, pp.33-40.

2. Lau, Antonio K.W (2007), “Educational

supply chain management: a case study”, Emerald

Group Publishing Limited, ISSN 1074-8121, Vol.

15 No.1, pp.15-27.

3. Habib, Mamun and Chamnong

Jungthirapanich (2008), “An integrated framework

for research and education supply chain for the

universities”, The 4th IEEE International

Conference on Management of Innovation and

Technology, Thailand.

4. Pathik, B. B., and Habib, M., “Redesigned

ITESCM Mode l: An Academic SCM for the

Universities”, International Journal of Supply Chain

Management (IJSCM), Vol. 1, Issue 1, pp. 12-26,

Excelling Tech Publishing Company, London, UK,

2012b. ISSN: 2050-7399 (online), 2051-3771 (print).

5. Pathik, B. B., and Habib, M., “Enhancing

supply chain management for the universities –

IESCM model perspective” International Journal

of Supply Chain Management (IJSCM), Vol. 1,

Issue 2, pp. 1-13, Excelling Tech Publishing

Company, London, UK, 2012c. ISSN: 2050-7399

(online), 2051-3771 (print).

6. University and Industry Relations in

Singapore. By Jasmine Kway, Deputy Director of

Industry and Technology Relations Office,

National University of Singapore.

TÓM TẮT

QUẢN TRỊ CHUỖI CUNG ỨNG GIÁO DỤC: GIẢI PHÁP NÂNG CAO

HIỆU QUẢ CÁC CHUYỂN GIAO KHOA HỌC CÔNG NGHỆ

Phạm Thị Minh Khuyên*, Phạm Thị Mai Yến Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên

Hoạt động quản trị chuỗi cung ứng giáo dục (IESCM) cho Trường cao đẳng và đại học cung cấp

hai đóng góp chính cho xã hội, bao gồm đóng góp về nguồn nhân lực và đóng góp về khoa học

công nghệ thông qua hai hoạt động chính là: đào tạo và nghiên cứu. Trên thế giới, lý thuyết về

chuỗi cung ứng giáo dục đã được nghiên cứu từ những năm cuối của thập niên 90, nhưng ở Việt

Nam đây vẫn là một nội dung rất mới. Bài viết này cung cấp một cái nhìn toàn diện, bao gồm các

yếu tố đầu vào, quá trình, và kết quả đầu ra của chuỗi cung ứng giáo dục. Với những phân tích về

ba nhóm hoạt động chính của quản lý chuỗi cung ứng và một số kinh nghiệm từ các nước có nền

giáo dục phát triển, bài báo đã làm rõ những ứng dụng của quản lý chuỗi cung ứng giáo dục đối

với các hoạt động trường đại học, đặc biệt là tăng cường hiệu quả của chuyển giao khoa học và

công nghệ.

Từ khóa: Quản trị chuỗi cung ứng giáo dục, Hiệu quả chuyển giao khoa học công nghệ, Quản trị mối

quan hệ, Định hướng xã hội, Chiến lược.


Thinh Duc Nguyen et al Journal of SCIENCE and TECHNOLOGY 127(13): 69 - 72

69

DESIGN AND FABRICATION OF ROBOTIC BLUETOOTH CLEANER

Thinh Duc Nguyen1,*, Thao Thi Phuong Phan1, Oanh Thi Lam Nguyen2

1University of Technology – TNU; 2Canon Vietnam .Co.Ltd

ABSTRACT The robotic bluetooth cleaner is a kind of the semi-automatic robot that can communicate with

smart devices such as smartphones, Computers by transferring bluetooth signal which is generated

in many electronic devices. This article presents a mechanical design and fabrication of the robot

as well as an algorithm for programming. The robot can move automatically on the floor and

collect dirt over a wide area by using triangular swept mechanism. The ultrasonic sensor detects

the distances from any obstacles to the robot then avoids them following the programmed paths.

This leading technology will dominate the modern domestics and vehicles in the near future.

Keywords: Bluetooth signal; robotic cleaner; domestic devices; untrasonic sensor.

STATEMENT PROBLEMS

Housework is now kind of tired and time-

consuming works. People clean by their

hands or some heavy wet cloth mops. It takes

lots of time to clearn a wide area.

Current products do not fulfill customer’s

requirements. There are various types of

cleaning machines such as vacuum cleaners

though people still have to do cleaning work

manually.

Additionally, science and technology are

developing quickly. People are desiring and

looking for a smaller product that can both

automatically and manually clean over the

wide area. This product can be connected to

smart phones or computers by a wireless

connection. And its price should be cheaper

than other common products.

Taking both benefits and drawbacks of

present designs, this project introduces an

engineering scope to develop a new design of

robotic cleaner.

CONCEPTUAL DESIGN

In this conceptual design the sensor detects

the distances from the obstacles to the robot

then the robot turns left or right atomatically

to avoid obtacles depending the program by

only one sensor. The zigzag path is

programmed to fulfill a wide area and two

driven motors controlled while swept

mechanism collecting all materials over a

wide area. It is shown on the Figure 1.

Furthermore the robot can stop automatic

running and change to manual control

alternately when the user want to swept dirts

over other areas.

Figure 1. 3D prototype of roboblue conceptual design on Solidwork software 2013*



70

Figure 2. The block diagram of algorithms

From the block diagram in Figure 2, a

bluetooth module is needed to let the robot

can go in two directions. Number 0 means no

bluetooth signal, so an Ultrasonic sensor is

necessary to detect the obstacle and measures

the distance (x) to the obstacle then defines

the order (a) of its in blue block. In case of

x<40cm & a=odd the roboblue turns left for a

certain time, in case of x<40cm & a=even the

roboblue turns right for a certain time, else it

moves forward. Number 1 is in case of that

the bluetooth module is receipt the signal

from the sources such as the laptop then

transfers it to the control center (main board

and driven board), so the roboblue can be

controlled manually such as turn left-right,

backward-forward or automatically.

ACTUAL DESIGN & CONSTRUCTION

The ping sent by ultrasonic sensor is equal

speed of sound 340 meters per second and the

sensor is only able to count time travelling

(microsecond), so to convert from time

measurement to distant measurement by using

the equation below we can count how long

the ping takes to travel 1cm:

(1)

So the distance from the robot to the obstacle

is counted by the equation below:

(2)

Where: t is read variable by the ultrasonic

sensor in microseconds.

Figure 3. The robotic bluetooth cleaner


71

From the idea of conceptual design, the actual

configuration is designed. A rectangular

shape make the robot smaller and stronger.

The swept system is design to parallel to the

floor with the distance 2mm (an optimized

distance to get as much dirt as possible). The

distance from upper face of swept mechanism

to robot frame is adjustable. The driven motor

of the swept mechanism is supplied by an

isolated circuit sourse. That hepls robot save

batery while it moves to other areas without

sweeping dirts. Its real picture is shown on

the Figure 3.

Program

The detail program is shown in the Apendix

Conclusion

The roboblue is absolutely following the

program and well operating. It can collect all

materials on its path such as dirt, pieces of

paper, hair pets, some particles like small

screws and nuts.

The roboblue operates on rechargeable

batteries. has lower product cost about 100

USD comparing with other product in the

same function.

Robotic bluetooth cleaners have been leading

the industry as domestic helpers for several

years now, with huge improvements in

functionality, performance and new state of

the art features. Simply schedule this robot to

clean the floors while you’re doing something

else.

Advancements in robotic bluetooth

technology have made impressive changes in

human ability to use smart robots and

motivated researchers developing science and

technology applications in human life.

REFERENCE

1. http://en.wikipedia.org/wiki/Vacuumcleanr

2. www.parallax.com/dl/docs/prod/acc/28015-

PING-v1.3.pdf.

3. http://www.genotronex.com.

Apendix const int trigPin = 2;

const int echoPin = 4;

int a = 0;

const int IN1=3;

const int IN2=5;

const int IN3=6;

const int IN4=9;

int SPEED_Control=200;

void setup() {

Serial.begin(9600);

pinMode( IN1 ,OUTPUT);




}

void loop()

{

long duration, inches, cm;

pinMode(trigPin, OUTPUT);

digitalWrite(trigPin, LOW);

delayMicroseconds(2);

digitalWrite(trigPin, HIGH);

delayMicroseconds(10);

digitalWrite(trigPin, LOW);

pinMode(echoPin, INPUT);

duration = pulseIn(echoPin, HIGH);

inches = microsecondsToInches(duration);

cm = microsecondsToCentimeters(duration);

delay(100);

if (cm < 40) {

a = a + 1;

Serial.print("a = ");

Serial.print(a);

Serial.print(" , ");

Serial.print(inches);

Serial.print("in, ");

Serial.print(cm);

Serial.print("cm");

Serial.println();

delay(10);

}

else {

a = a;

Serial.print("a = ");

Serial.print(a);

Serial.print(" , ");

Serial.print(inches);

Serial.print("in, ");


72

Serial.print(cm);

Serial.print("cm");

Serial.println();

delay(10);

}

if (cm <40 && (a%2)){

LEFT(200);

delay(10000);

}

else if (cm < 40 ) {

RIGHT(200);

delay(10000);

}

else {

FORWARD(250);

delay(10); }

}

long microsecondsToInches(long microseconds)

{

return microseconds / 74 / 2;

}

long microsecondsToCentimeters(long

microseconds)

{

return microseconds / 29 / 2;

}

analogWrite(IN1,0);

analogWrite(IN2,Speed);

analogWrite(IN3,0);


void FORWARD(int Speed){

}

void BACKWARD(int Speed){


analogWrite(IN2,0);


analogWrite(IN4,0);

}

void LEFT(int Speed){

analogWrite(IN1,0);


analogWrite(IN3,0);

analogWrite(IN4,0);

}

void RIGHT(int Speed){

analogWrite(IN1,0);

analogWrite(IN2,0);

analogWrite(IN3,0);


}

void STOP(){

analogWrite(IN1,0);

analogWrite(IN2,0);

analogWrite(IN3,0);

analogWrite(IN4,0);

}

TÓM TẮT

THIẾT KẾ VÀ CHẾ TẠO QUÉT NHÀ THÔNG MINH

Nguyễn Đức Thịnh1,*, Phan Thị Phương Thảo1, Nguyễn Thị Lâm Oanh2 1Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên;

2Công ty TNHH Canon Việt Nam

Nghiên cứu xây dựng và chế tạo một Robot quét nhà thông minh, có thể được điều khiển thông

qua tín hiệu bluetooth bằng điện thoại smart phone hay máy tính có thiết bị phát bluetooth hoặc

robot có thể tự động quét nhà theo một chương trình được thiết lập trước. Robot dùng cảm biến

siêu âm để phát hiện vật cản và đo khoảng cách từ nó tới robot từ đó truyền tín hiệu tới mạch điều

để thay đổi hướng di chuyển của robot. Cơ cấu chổi quét ba trục có thể thay đổi được khoảng các

tới mặt sàn quét tôi ưu trong việc gom rác trên mặt sàn. ông nghệ điều khiển thiết bị bằng tín hiệu

bluetooth này trong tương lai sẽ được áp dụng rộng rãi cho điều khiển các hệ thống thông minh

như trong ô tô, trong các ngôi nhà thông minh.

Từ khóa: Tín hiệu Bluetooth, robot quét nhà, thiết bị trong nhà, cảm biến siêu âm.



73

CACULATION ANALYSIS AND DESIGN FOR CONSTRUCTION OF SOLAR ENGINE MODEL

Thinh Duc Nguyen*, Tuan Anh Vu, Du Van Nguyen


ABSTRACT

This research constructs a mathematical model of Stirling engines on Matlab, bases on the given

input as the power, speed, temperature, working gas, power loss to figure out the most suitable

design for construction of the engine that can reach the highest efficiency, simplify structure, and

easily operate. Then designing and simulating dynamic operation of physical model of the engine

on Inventor. Finally a real engine was manufactured and assemblied.

Keywords: Stirling engine; solar energy; mathematic model.

INTRODUCTION*

While fossil fuels are getting empty and every

nation has to face extreme influences due to

the global warming phenomena, it has raised

great demands on available replacements of

recycle and green energies in this decade.

Applications of stirling engine open a trend to

solve the existing problems. It is a kind of

external combustion engine using green

source and friendly with environment;

therefore, successful implementation of

stirling engines in social life resolves

environmental problems and contributes

much for a national economy.

Up to now, there have been any detailed and

successful researches on design and

production of stirling engines in Vietnam.

Early in 1983, Professor Ivo Kolin of the

University of Zagreb, Croatia, demonstrated

the very first low temperature differential

Stirling engine to an amazed audience. In

1992 Professor Senft was asked to design and

build a low temperature differential engine for

NASA. This engine, called the N-92, was

optimized for hand held operation, with a

temperature difference as low as 6°C enough

to power it. This reseach focused on

studying different types of stirling engines

which could be beneficial applications in the

local conditions such as solar water pump,

generator. [1] This paper presents our

* Email: [email protected]; Tel: 0946650662

successful work on design and optimization

process on the engine manufacturing with any

given inputs.

OPERATION

The working fluid is a gas operating under

relatively high temperatures and obeys the gas

laws. When the gas is heated and because it is

in a sealed chamber, the pressure rises and the

force due to its pressure acts on the power

piston to produce a power stroke.

When the gas is cooled the pressure drops and

this means that less work needs to be done by

the piston to compress the gas on the return

stroke, giving a net gain in power available on

the shaft.

The efficiency of any Thermodynamic cycle

increases as the average temperature of heat

addition increases. Regenerators are used for

this purpose [2].

Figure 1. The ideal Stirling Cycle

in PV and TS diagrams

Process 3– 4, isothermal expansion process.

Process 4–1, constant volume regenerative

transfer process.


74

Process 1–2, isothermal compression process.

Process 2–3, constant volume regenerative

transfer process. [3]

3D MODEL DRAWN ON INVENTOR &

MATHEMATICAL MODEL

PROGRAMMED ON MATLAB

3D model drawn on Inventor:

Since the theory and operation of the ideal

Stirling Cycle in diagrams, 3D model of

gamma type stirling engine was buit. Three

were essemplied perpendicular to each other

that made the engine strong and stable. All

elements were designed on Inventor software.

Figure 2. 3D mode of Stiring engine

The engine included five main parts:

1. Heater.

2. Heat exchanger.

3. Power piston.

4. Frame.

5. Crank-shaft mechanism.

The engine was dynamically simulated well

on the software.

Mathematical model programmed on

Matlab:

Input: Temperature; speed; working gas

volume; power. [4] design parameters were

investigated to find out the optimistic design

by Matlab programming based on mathematic

equations below.

a, Total volume:

C R EV V V V (1)

Where: EV is expansion volume.

2

SEE

VV (1- cos x ) + VDE (2)

RV is generator volume.

CV is compression volume.

1 os2

SCC DC

VV c x dx V

(3)

With:

VSE is swept volume of the expansion piston.

VDE is expansion dead volume.

VSC is swept volume of the compression

piston.

VDC is a compression dead volume.

x - is crank angle. x=0 when the expansion

piston is located the most top position (top

dead point).

dx - is phase angle.

b, Engine pressure is calculated by:

2 2 21

cos 1 .cos

Pmean S B Pmean cP

S B x a c x a

(4)

With : min

1

1

mean

PP

c

c

(5)

(Pmin = 101.3*103Pa)

1 sintan

cos

v dxa

t dx

(6)

42 2

1

EDE DC

tXS t tX v X

t

(7)

2 22 cosB t tv dx v (8)

Bc

S (9)

With dimensionless parameter:

C

E

Tt

T (Tc,Te – compression and expansion

space gas temperature) (10)

SC

SE

VV

V (11)

DEDE

SE

VX

V (12)


75

DCDC

SE

VX

V (13)

RR

SE

VX

V (14)

c, Output power is calculated by:

W W60

e c

nLi

With:

2

. . . .sinW

1 1

mean SE

e

P V c d a

c

(15)

2

. . . . .sinW

1 1

mean SE

c

P V c t d a

c

(16)

RESULTS

- Result from Matlab program in the fig 3.

When the design parameters were changed,

the program would caculate the power and

plot it in the figure on the screen.

The results from Matlab program are recorded

on Excel in the fig. 4.

The result from the Figure 4 shows that the

design parameters on the fourth line of the

table are the most suitabe for the construction

of the engine. Output power is max at

TC = 300K, TE = 600K, n = 500rpm, with

VR = VDC = VDE = 10cm3, VC = VE =

80cm3 in our limited investigate range.

As the results from the investigation above, a

real model was designed. Its real picture can

be seen in the Figure 5.

CONCLUSION

With the simple in design and high output

power, once time stirling engine α-type shows

its dominant comparing with the others. This

research helps designers easily approach

optimum in manufacture and save time of

designing.

Specially, it is not harm to the environment.

Since 1990s Molten salt storage has been

discover which is a mixture of 60 percent

sodium nitrate and 40 percent potassium-

nitrate, commonly called saltpeter retains

thermal energy collected for later uses.They

can be employed to balance energy demand

between day time and night time. So solar

engine is a kind of the expected in the future.

REFERENCE

1. C. Julian Chan, Physics of solar energy, Wiley

publisher, 1 edition July 26, 2011

2. G Venugopal, Stirling Engines - A Beginners

Guide, revised version-2012, PDF.

3. http://en.wikipedia.org/wiki/Stirling_engine

4. http://www.bekkoame.ne.jp/~khirata/acad

Figure 3. The result from Matlab program


76

Figure 4: Recorded results on Excel

Figure 5. The real stirling engine

TÓM TẮT THIẾT KẾ VÀ CHẾ TẠO ĐỘNG CƠ NHIỆT STIRLING SỬ DỤNG NĂNG LƯỢNG MẶT TRỜI

Nguyễn Đức Thịnh*, Vũ Anh Tuấn, Nguyễn Văn Dự


Nghiên cứu xây dựng một mô hình toán học của động cơ nhiệt Stirling trên phần mềm Matlab dựa

vào các thông số cho trước như công suất đầu ra, tốc độ vòng quay, nhiệt độ cấp vào, khí làm việc,

năng lượng mất đi để tìm ra một kết cấu của động cơ phù hợp nhất, đạt hiệu suất cao nhất, đơn

giản hóa kết cấu và vận hành một cách dễ dàng. Sau đó thiết kế và mô phỏng chuyển động của mô

hình vật lý của động cơ trên phần mềm Inventor rồi đi đến gia công chế tạo một mô hình động cơ

thật vận hành dựa vào nguồn năng lượng nhiệt được cấp từ ánh sáng mặt trời.

Từ khóa: Động cơ Stirling, năng lượng mặt trời, mô hình toán học.


Ha Thi Thu Phan et al Journal of SCIENCE and TECHNOLOGY 127(13): 77 - 80

77

EFFECT OF ANNEALING TREATMENT ON HIGH STRAIN RATE BEHAVIOR OF GRAPHENE REINFORCED POLYURETHANE COMPOSITES

Ha Thi Thu Phan, Thao Thi Phuong Phan


ABSTRACT Research efforts in mechanical behaviors at high strain rates in specific ranges on the post-heated

Graphene reinforced polyurethane composite continue to grow in this paper. Unheated and post-

heated sample types with different Graphene concentrations (0.25%wt, 0.5%wt, 0.75%wt and

1%wt) have been tested at high strain rate regimes of 1500 s-1 to 5000 s-1 to see if being heated at

55C in 12 hours affects on high strain rate behavior of the composite. Results show that the post-

heated composite is a strong strain rate dependent material. Stress-to-break of 200 Mpa to 270

Mpa of post-heated composites is greater than stress-to-break of around 150 Mpa of unheated

composites.

Keywords: Graphene, Polyurethane, high strain rate, annealing.

INTRODUCTION*

The Graphene (GR) reinforced Polyurethane

(PU) composite (GR/PU), a new nano-

composite material fabricated by GR fillers

and PU matrix, has drawn considerable

attention due to its unique and outstanding

mechanical, electrical, and thermal properties.

The study on the dynamic behavior of the

composite at high strain rates [1] makes

available data using in the composite’s

applications such as flexible packing, semi-

conductive sheets in transistors, memory

devices, hydrogen storage, electronics

(sensors), etc.

Elastomer PU consist of chains organic units

joined by urethane links. Its components are

divided into two phases: hard segment (HS)

and soft segment (SS). At room temperature,

HS has high density and is rigid, while SS has

very low density and is flexible. SS forms an

elastomeric matrix responsible for elastic

properties of PU, and HS acts as

multifunctional tie points, functioning as both

cross-links and reinforcing fillers. Although

GR is considered as an excellent thermal

sustenance material with thermal conductivity

of 5000 W/m.K and melting temperature

about 3000C [3]. Therefore, at conventional

annealing temperature, it is not easy for GR to


be affected, but the property of GR/PU

composite may change after annealing

treatment due to PU’s relatively low melting

temperature. Property of PU may change

when only SS melts while HS is remained,

leading to property change in GR/PU

composite. Base on the knowledge that the SS

melting temperature of most PU types is

lower than 50C, this study conducted

experiments with GR/PU composite specimen

which is annealed at 55C in 12 hours to see

if any change exists in its dynamic

mechanical property compared to that of

unheated specimens.

EXPERIMENTAL STUDIES

GR/PU composite is prepared by solution

mixing method [2]. GR was added into 20-40

ml acetone, and then agitated its particles by

sonication. The mixture of acetone and GR

was poured into PU solution and sonicated it.

Acetone, then, was come out of the mixture

by vacuuming. Heat is necessary if acetone is

hard to get rid of, but the temperature should

not be too high. The shape of the composite

plate was molded by a spacer after three days.

For, post-heated GR/PU composite, it was

annealed at 55C for about 12 hours. Square

specimens with ¼ in. (6.35 mm) side length

and 1/8 in. (3.175 mm) thickness were cut and

polished on two impact sides by sand papers.


78

A compressive Split Hopkinson Pressure bar

(SHPB) is used to investigate the dynamic

mechanical behaviors of GR-reinforced-PU

composites (GR/PU) at high strain rates ranging

from approximately 1500 s-1 to 5000 s-1.


Similar to stress-strain curves of unheated

specimens, stress-strain behavior of post-

heated specimens are characterized by three

distinct regions as shown in Fig.1: 1) Elastic

region in which stress and strain are linearly

related, 2) plateau region where stress is

almost constant while strain keeps increasing,

and 3) strain-hardening region in which stress

increases with strain. After reaching peak

stress (the summit point of strain hardening

region), stress decreases due to unloading.

Both unheated and post-heated specimens,

peak stresses and peak strains are strongly

dependent on strain rates, especially in a

strain rate range of 3000 s-1 to nearly 5000 s-1.

Among pristine PU and GR/PU composites

with different weight percentages of GR, the

composite of 0.5% wt. GR shows good

increasing trend of peak stresses compared to

the rest, especially with strain rates higher

than 3000 s-1. It is a result ofideal strain

hardening well documented in [1].

a/ b/

Figure 1. Dynamic compressive stress-strain responses of GR/PU composites with 0.75%wt.GR;

a/ Unheated composites; b/ Post-heated composites

a/ b/

Figure 2. Strain rate vs. peak stress of GR/PU composites in different GR contents:



79

Strain rate ranging from 3000 s-1 to 4000 s-1,

peak stress of a post-heated specimen is

greater than unheated specimen’s. For

example, considering the composite with

1%wt GR, at strain rate of 3600 s-1, peak

stress of unheated specimen is 180 Mpa,

while post-heated specimen obtains the value

of 140 Mpa. In a range of 4000 s-1 to 5000 s-1,

most unheated specimens are broken at a

strain rate higher than 4000 s-1, but post-

heated specimens remain until nearly 5000 s-1

and peak stresses keep growing. As a result,

stress-to-break, a peak stress value at which

the specimen is broken, of post-heated

specimen strongly is higher than unheated

composite with the same GR content. For

instance, stress-to-break of the post-heated

composite with 1%wt. GR is about 230 Mpa,

while the value of unheated corresponding

composite is 150 Mpa.

The relation between peak strain and strain

rate are shown in Fig. 3. Though the shapes of

plots are similar, post-heated specimens

present higher strain-to-breaks at higher strain

rates compared with unheated composites.

To explain for the phenomenon pointed out

above, melting temperature of SS and HS

plays a key role. When annealed at 55C for

12 hours, only SS experiences melting

process and it becomes melting viscous. If

any void exits in SS during composite-making

process, it may disappear. After melting, SS is

more homogenous and attainable due to its

crystallinity at high loadings as the composite

recovers again two-phase (SS and HS)

properties, but greater uniformity in SS.

Additionally, annealing GR/PU composite

makes stronger hydroxyl, so the interaction

between hydroxyl of GR and diisocyanate of

PU is stronger [7]. As a result, these strong

interactions increase the mechanical

properties of the composite.

Another reason for the added composite

strength may be attributed to PU matrix. PU,

as introduced, is a so-called “self-reinforced”

polymer because of hard segments (HS)

which both play as cross-link and reinforcing

fillers. After annealing, though HS is not

melting at the temperature of 55C, it extends

and interacts more strongly, giving rise to

good dynamic properties [5, 8].

CONCLUSIONS

All tested specimens including unheated and

post-heated GR/PU composites are observed

to be highly strain rate sensitive. However,

strain rate that the post-heated composite can

suffer before broken is 1000 s-1 higher than

unheated composite’s.

Stress-to-break of 200 Mpa to 270 Mpa of the

post-heated composite is greater than that of

unheated composite which just obtain the

value lower than 150 Mpa. Strain-to-breaks of

unheated and post-heated composites are not

much dissimilar. On plots, this different value

is about 0.01 m/m.

a/ b/

Figure 3. Strain rate vs. peak strain of GR/PU composites in different GR contents:



80

REFERENCES

1. Ha Phan and The Phan, “Mechanical behavior

at high strain rates of Graphene reinforced

polyurethane composites with different Graphene

volume fraction”, ISEPD, vol.1, pp.56-60, 2014

2. Xin Wang, Yuan Hu, Lei Song, Hongyu Yang,

Weiyi Xing and Hongdian Lu, “In situ

Polymerization of Graphene nanosheets and

Polyurethane with enhanced mechanical and

thermal properties”, Journal of Materials

Chemistry, vol. 21, pp. 4222-4227, 2011

3. Tapas Kuilla, Sambhu Bhadra, Dahu Yao,

Nam Hoon Kim, Saswata Bose, Joong Hee Lee,

“Recent advances in Graphene based polymer

composites”, Progressive in Polymer Science, vol.

35, pp. 1350-1375, 2010.

4. Bazle A. Gama, Sergey L. Lopatnikov, John

W. Gillespie Jr., “Hopkinson bar experimental

tenique: A critical review”, American Society of

Mechanical Engineers, vol. 57, No. 4, pp. 223-

250, July 2004.

5. Yingjie Li, Tong Gao, Jian Liu, Kung Linliu,

C. Richard Desper, and Benjamin Chu,

“Multiphase structure of segmented PU: Effects of

Temperature and Annealing”, Macromolecules,

vol. 25, pp. 7356-7372, 1992.

6. Fengdan Jiang, Liqun Zhang, Yi Jiang,

Yonglai Lu and Wencai Wang, “Effect of

Annealing treatment on the structure and

properties of Polyurethane/Multiwalled carbon

nanotube nanocomposites”, Journal of Applied

Polymer Science, pp. 1-8, 2012.

7. Cristina Prisacariu, “Polyurethane Elastomers:

From Morphology to Mechanical Aspects”,

Springer, 2011.

8. Fengdan Jiang, Liqun Zhang, Yi Jiang,

Yonglai Lu and Wencai Wang, “Effect of

Annealing treatment on the structure and

properties of Polyurethane/Multiwalled carbon

nanotube nanocomposites”, Journal of Applied

Polymer Science, pp. 1-8, 2012.

TÓM TẮT

ẢNH HƯỞNG CỦA NHIỆT ĐỘ ĐẾN ĐẶC TÍNH BIẾN DẠNG TỐC ĐỘ

CAO CỦA VẬT LIỆU COMPOSITE GRAPHENE NỀN POLYURETHANE

Phan Thị Thu Hà*, Phan Thị Phương Thảo


Những nghiên cứu về đặc tính cơ học ở một số dải biến dạng tốc độ cao của vật liệu composite

Graphene nền Polyurethane tiếp tục được đề cập đến trong bài báo này. Các mẫu thử chưa qua lò

nung và đã qua lò nung với phần trăm khối lượng Graphene khác nhau (0.25%GR; 0.5%GR;

0.75%GR và 1%GR) được thử ở dải tốc độ biến dạng cao 1500 1/s đến 5000 1/s để thấy được ảnh

hưởng của việc nung nóng mẫu thử đến 55 độ C trong 12 giờ ảnh hưởng thế nào đến cơ tính của vật

liệu composite. Kết quả chỉ ra rằng mẫu thử vật liệu đã qua nung nóng rất phụ thuộc vào tốc độ biến

dạng. Ứng suất đạt được trước khi mẫu đã qua nung bị phá hủy là khoảng từ 200 MPa đến 270 MPa,

trong khi đó ứng suất đạt được trong trường hợp mẫu chưa qua nung chỉ là khoảng 150 MPa.

Từ khóa: Graphene, Polyurethane, biến dạng tốc độ cao, nung nóng.


Thuy Thi Hong Truong et al Journal of SCIENCE and TECHNOLOGY 127(13): 81 - 86

81

NEURAL NETWORK APPLICATION TO THE DIAGNOSIS OF HEPATITIS

Thuy Thi Hong Truong*, Nga Thi Hong Do

University of Medicine and Pharmacy – TNU

ABSTRACT Neural network can be used in many different problems that existed in the relationship of input

and output. One of the biggest advantages of neural network is that it is able to solve the problems

which have no algorithm or too complicated algorithms. Problem diagnosis in medicine is a good

example. This paper has studied the application of neural networks in medical diagnosis, some

problems of building a decision support system diagnosis and testing neural networkʼs application

to hepatitis diagnosis based on training with Wiscosin hepatitis data, to provide an overview of

possible applications of powerful information technology in the field of medicine.

Keyworks: Neural Network, Medical diagnosis, Decision support system, Training, Wiscosin

hepatitis data.

INTRODUCING THE NEURAL NETWORK MODEL IN DIAGNOSIS*

Most physicians have diagnosed diseases based on the knowledge accumulated in

colleges, training courses, ... However, the medical knowledge is often quickly out of

date, in order to diagnose, doctors must have enough experience (for about 10-20 working

years). Besides, doctors also have difficulty in diagnosing rare diseases or emerging

diseases. The solution can be used to bring the benefits of computers to improve the

diagnostic capabilities such as: using the data collected from experienced colleagues, they

have been using the experience gathered owing to connecting to the world,... [4] so far,

neural networks have achieved many remarkable achievements when applying to

many different areas of medicine such as disease diagnosis, medical image analysis,

biomedical analysis,...

Neural network applications in medical

diagnosis support is, in fact, essentially solving the problem of classification of

medical statistics. The biostatistics data are often given in tabular format in which each

line is a record, each column is a symptom and a column is to determine the diagnosis.

The input of the neural network will be the first symptom and the output will be the

diagnosis.

* Email: [email protected]

SEVERAL ISSUES FOR DEVELOPING

DIAGNOSIS SUPPORT SYSTEMS

To build networks, which are capable of

decision support and have diagnostic

accuracy and high performance, we have to

solve some of the following issues:

Data preprocessing

This phase plays an important role in the

process of building systems by data sets in the

individual studies, which are often too small

to produce reliable results. Besides, the data

entry process and measurement errors also

appear due to typing error or not checking the

boundary conditions of the variables...

Therefore, we need to better define the

characteristic variables, remove or fix

incorrect data to build reliable systems.

According to experts, the number of neural in

hidden layer affects the generalization ability

of the network. If a neural network has a

small number of hidden layers, that will not

be defined, the structure of the network to

perform adequately in the training data, in this

case, it is considered to be unsuitable

(underfitting). Conversely, if the number of

hidden layer neural network is too big, the

network can not clearly define the decision

boundary in the vector space which is

affected by the level set by the properties of

the training data network. In this case, it is

considered to be too joints (overfitting). The


82

question is how much neural selected in the

hidden layer is appropriate? We can use

methods of learning increasingly in

progression of Kavzoglu to determine the

optimum values for the number of neural in

hidden layer:

1. Select an initial value for the number of

neural in hidden layers.

2. Repeat the following until it reaches a

predetermined value, which is the maximum

number of neural requiring survey:

- Train the neural network with several

hidden layers present. Record network

training time.

- Check the accuracy of the network with test

data sets, respectively.

- Increase the number of hidden layer neural 1.

The illustration shows the error rate and

training time corresponding the neural

network in the hidden layer. This network

model has high accuracy corresponding to the

network training time that is acceptable to the

number of neural in hidden layer between 10

and 15 neural. And when the number of

neural in hidden layer reaches 15, the network

has the highest accuracy (about 92%), then

we will choose the network with 15 hidden

nodes in this case [1].

Training and assessment, network model

selection

Network training process is to find the

weights of network so that mapping from the

data entering into the domain of the output

value is matched with the set pattern. Some

rules in neural network training: Reducing

Gradient, studying a sample of each school,

the moment ...

In addition to the selection of rules with

proper training, the selection of network

model with parameters appropriate training

network has much influence on the results

obtained by the neural network. In fact, to

evaluate the ability of the model subclass N

network based on D data set, we can use

several methods: Hold-out method, K-fold

method, the method bootstrap methods

divided into three data lines... We can do by

three-way data spliting to select the network

model and correct the error assessment that

divides the data into three distinguishable

data sets:

Training set: The training data of the network

to match the parameters of the network. In the

MLP network, it is used to self centimeters

optimal weights for back propagation rule.

Figere 1.Defect rate and training time to the number of neural in hidden layer

Stu

dy

Tim

e (s

) P

reci

sio

n

The number of neural in hidden layer.

Rate failed.

Training period.


83

Validation set: The data set is used to refine the parameters of the network. In the MLP network, it is used to determine the number of neural in hidden layers of network or define stops for back-propagation algorithm.

Testing set: The data is used to evaluate the performance of a full network. In the MLP network, it is used to evaluate the error rate of the network after selecting the final model.

The steps taken by the method of three-way data sharing is done following:

Step 1: Divide the data set into three distinct subsets: training, validating and Testing.

Step 2: Select the network architecture and the network training parameters.

Step 3: Train the network with the training set.

Step 4: Use the validation set to evaluate the model.

Step 5: Repeat steps 2,3,4 with the architecture and different network training parameters.

Step 6: Choose the best network model and train in both training and validation data sets.

Step 7: Use the test set to evaluate the final network model.

NEURAL NETWORK APPLICATION TO THE DIAGNOSIS OF HEPATITIS

Hepatitis is a very dangerous disease because it is silently destroying livers which patients have little or no obvious symptoms until disease becomes severe and high mortality can happen. Especially, this disease can be treated completely if it’s diagnosed early. So in the process of diagnosis, identifying patients at risk of death or not is very important. We can use a neural network in diagnosis based on a database to solve subclass. The data set is in tabular format, where each line corresponds to a record, each column corresponds to a symptom, one column to determine results.

To diagnose hepatitis, we can use a database of hospital hepatitis University of Wisconsin (USA) with 18 million certified inputs at integers or real numbers:

To design a hepatitis diagnosis system, we can use Alyuda Neurointelligent program.

This is a software to support experts to solve the problem of forecasting ... When we do we have to solve the following issues:

(1)Age (0-100)

(2)SEX: male, female (1,2)

(3)STEROID: no, yes (1,2)

(4) ANTIVIRALS: no, yes (1,2)

(5) FATIGUE: no, yes(1,2)

(6) MALAISE: no, yes(1,2)

(7) ANOREXIA: no, yes(1,2)

(8) LIVER BIG: no, yes(1,2)

(9) LIVER FIRM: no, yes(1,2)

(10) SPLEEN PALPABLE: no, yes(1,2)

(11) SPIDERS: no, yes(1,2)

(12) ASCITES: no, yes(1,2)

(13) VARICES: no, yes(1,2)

(14)BILIRUBIN:(0-10)

(15)Alkaline phosphatase:(0-500)

(16) SGOT: (0-1000)

(17) ALBUMIN: (0-10)

(18) HISTOLOGY: no, yes (1,2)

In addition to 18 values above, using one determined value is DIE (1), LIVE (2).

The number of samples

With a database obtained from the University of Wisconsin Hospital policy contains 155 acres, which are divided into two layers, deaths have 32 samples and living patterns have 123 samples.

Analysis of data

The data set contains some faulty attribute values, denoted by a "?". Stored in data file viemgan.csv. Run the program and open the file Alyuda Neurointelligent viemgan.csv. Then the program will automatically adjust the data analysis omitted values in the data set and data divided into 3 parts:

- Training set: The data set is used to train the network.

- Validation set: The data set is used to refine structure or other parameters of the network instead of using neural networks.

- Test set: The examined data set. Default accounts for 15% of the sample.


84

Table 1. The patient data samples

61 1 1 2 1 2 2 1 1 2 2 2 2 1.3 78 25 3.8 100 1 2

51 1 1 1 1 1 2 2 2 2 2 2 2 1 78 58 4.6 52 1 2

39 1 1 1 1 1 2 2 1 2 2 2 2 2.3 280 98 3.8 40 1 1

62 1 1 2 1 1 2 ? ? 2 2 2 2 1 ? 60 ? ? 1 1

41 2 2 1 1 1 1 2 2 2 2 2 2 0.7 81 53 5 74 1 2

26 2 1 2 2 2 2 2 1 2 2 2 2 0.5 135 29 3.8 60 1 2

23 1 2 2 1 1 1 2 2 1 2 2 2 1.3 194 150 4.1 90 1 2

20 2 1 2 1 1 1 1 1 1 1 2 2 2.3 150 68 3.9 ? 1 2

42 1 1 2 2 2 2 2 2 2 2 2 2 1 85 14 4 100 1 2

65 1 2 2 1 1 2 2 1 1 1 1 2 0.3 180 53 2.9 74 2 2

Before putting some values into the neural network, all of them need to be scanned and

converted because in the neural network, numerical values only work in a certain range.

In the first column the scan range is [-1 .. 1], In the purpose column, this range depends on

the activation function on the output layer. To scan and convert numeric values, we use the

formula:

Scanning

coefficient:minmax

minmax

XX

SRSRSF

Values after treatment:

XP = Srmin + (X - Xmin)*SF

Among them:

X: actual value of a numeric column

Xmin: minimum actual value of the column

Xmax: maximum actual value of the column

SRmin: lower scaling range limit

SRmax: upper scaling range limit

SF - scaling factor

Xp - preprocessed value

Network Design

To design a network, we need to determine the network architecture (the number of

hidden layers and units in each layer) and the network properties (error and activation

function). The properties of the network can be defined automatically but you can change

them manually, but in some cases, to improve network performance. To classify problems

you can also choose a classification model and its parameters.

Figure 2. The optimal neural network


85

Figure 3. Diagnosis Query

Training Network

Implementation of network training on the

number of layers in the neural network and

hidden layer was designed by network

training algorithms.

After training, we can know the accuracy

achieved on the training data set, the training

process loop stops at many things, network

training speed... Basing on this result, we can

change the network parameters or return to

coaching network designed to select an

optimal network model.

Test results of network training

Testing selection on the entire data set. The

test results showed that the diagnosis gets

high achievement (> 90%).

After network construction is complete, we

can conduct using the network to diagnose in

the query.

CONCLUSION

The diagnosis of the doctor is not always

accurate, especially of young doctors who

have little experience of rare or new diseases.

Therefore, the development of a diagnosis

decision support system brings many practical

implications. It not only supports clinicians to

reinforce the accuracy of its decision, but also

brings experience in the diagnosis based on

the number of samples.

REFERENCES

1. Hongmei Yana, Yingtao Jiangb, Jun Zhenge,

Chenglin Pengc, Qinghui Lid, (2006), “A

multilayer perceptron-based medical decision

support system for heart disease diagnosis”,China.

2. J. B. Siddharth Jonathan and K.N. Shruthi,

(2002), “Two tier Neural Inter-network based

Medical Diagnosis using k-Nearest Neighbor

Classification for Diagnosis pruning”,USA.

3. P. Venkatesan, S. Anitha, (2006). “Application

of a radial basis function neural network for

diagnosis of diabetes mellitus”. Tuberculosis

Research Centre, ICMR, Chennai 600 031, India.

4. Rüdiger W. Brause, (2006). “Medical Analysis

and Diagnosis by Neural Networks”. J.W. Goethe-

University, Computer Science Dept., Frankfurt a.

M., Germany.

5. Tüba Kiyan, Tülay Yildirim, (2004). “Breast

Cancer Diagnosis Using Statistical Neural

Networks”. Department Yildiz Technical

University Besiktas, Turkey.


86

TÓM TẮT

ỨNG DỤNG CỦA MẠNG NEURAL

TRONG CHẨN ĐOÁN BỆNH VIÊM GAN

Trương Thị Hồng Thúy*, Đỗ Thị Hồng Nga

Trường Đại học Y Dược – ĐH Thái Nguyên

Mạng neural có thể được sử dụng trong nhiều bài toán khác nhau mà tồn tại các mối liên hệ giữa

đầu vào và đầu ra. Một trong những ưu điểm lớn nhất của mạng neural là có khả năng giải quyết

các bài toán không có thuật toán giải hoặc thuật toán giải của nó quá phức tạp. Bài toán chẩn đoán

bệnh trong y học là một ví dụ điển hình. Bài báo này đã tiến hành nghiên cứu ứng dụng của mạng

neural trong chẩn đoán bệnh, một số vấn đề khi xây dựng một hệ hỗ trợ ra quyết định chẩn đoán

bệnh và thử nghiệm ứng dụng của mạng neural để chẩn đoán bệnh viêm gan dựa trên việc huấn

luyện với số liệu mẫu của đại học Wiscosin, nhằm cung cấp một cách nhìn tổng quát về khả năng

ứng dụng mạnh mẽ của công nghệ thông tin trong lĩnh vực y học.

Từ khóa: Mạng neural, chẩn đoán y học, hệ hỗ trợ ra quyết định, huấn luyện, dữ liệu viêm gan

Wiscosin.

* Email: [email protected]

Huy Ngoc Vu et al Journal of SCIENCE and TECHNOLOGY 127(13): 87 - 91

87

THYRISTOR-BASED DIGITAL CONTROL OF DC MOTORS

Huy Ngoc Vu*, Tuan Manh Tran, Huong Thi Mai Nguyen, Hung Tien Nguyen University of Technology – TNU

ABSTRACT DC motors have been widely used in the industry because of high mechanical power density,

simplicity and cost effectiveness as well as of their versatile control characteristics. Speed of the DC

motor can be controlled by regulating its armature voltage using some of the thyristor-based circuits.

In the past, the speed control of DC motors was accomplished with analog technology due to simple,

and low cost constraints. However, this technique exhibits many drawbacks since it can not guarantee

the drift of device parameters in a desired range when working conditions are changed. This paper

proposed the speed control of DC motors using a digital signal processor (DSP) which has the special

features for digital motor control. The velocity control loop uses PI controller that has been

implemented by programming in the DSP core. According to the input command, feedback command

and control algorithm, the PWM pulses generated by the DSP are given to gates of thyristors.

Keywords: DC motor, digital signal processing, DSP, PWM.

INTRODUCTION*

Currently in DC motor drive systems are still

widely used because it has the following

advantages: High starting torque, ability to

adjust the speed is relatively simple, even for

the high power motor, easy to reverse

rotation, stabilizing the speed easily, the

structures of the power circuit and the control

circuit are relatively simple, can achieve high

adjustable quality in wide speed range.

Most of the drive systems using DC motors

are required to adjust the speed. In fact there

are two basic methods to adjust the speed of

DC motors: Adjustable voltage for the motor

armature and adjustable voltage for motor

excitation circuit.

The controlled rectifier is in the power circuit

of the drive system to adjust the motor speed.

These converters’ outputs apply to the motor

armature circuit or motor excitation circuit.

The converters, the circuits and the control

algorithms can be implemented using analog

or digital techniques. In particular, the use of

analog techniques has the advantage of being

relatively simple, efficient and has developed

completely. Basic drawback of this technique

depends heavily on parameters’ drift of the

elements in the system making the stability of


the system sometimes not be guaranteed over

time and be difficult to response to the

different working conditions.

The drive system of the DC motor typically

uses two adjustable loops. In the outer loop is

the velocity regulation, the inside loop is the

current regulation. Current regulation loop

required to response faster than the velocity

regulation loop that depends heavily on

mechanical inertia of the motor and the load.

Therefore, the design requirements for the

current loop have more stringent

requirements. To get the optimal parameters

for the current regulation loop, there must be

the exact parameters of the motor. Then, the

controller in analog technique of the current

loop is adjusted according to the parameters

have been synthesized by changing the value

of the variable resistor, variable capacity or

the gain of the amplifiers. However, in the

operation, the parameters of the motor may be

altered due to their dependences on working

conditions (resistance of the armature varies

with temperature, inertia varies with the

load...), the electronic components themselves

as well as operational amplifiers also changes

with temperature, humidity... Therefore, the

quality of the controller in analog technique is

very difficult to ensure over time and the

different working modes.


88

Digital control techniques are not only

capable of overcoming the aforementioned

drawbacks of the analog control systems but

also open up the possibility to easily apply

modern techniques in the synthesis of

controllers easy to change the operating

parameters of the devices, capable of

interacting with humans, the ability to pair

with the computer, the control devices of

higher level or the other digital interfaces (for

example the programmable logic controller -

PLC, the digital speed measurement unit

Encoder)... Furthermore, a digital control

system also allows removal of a significant

number of analog circuits having different

functions to perform by software (eg,

measurement circuits, protecting,

displaying...) make the controller more

compact and reliable. Also, in reality the

digital speed controllers for DC motors in

Vietnam are offered by most famous brands,

high cost, especially in large power range,

complex warranties and maintenances.

Therefore, the research and development of

digital control system of a DC motor speed

can improve the ability to master the

technology, improve the quality of

equipments and facilities for the use and

operation, repair and reduce costs.

THE MATHEMATIC MODEL OF DC

MOTOR

The voltage equivalent equation of DC motor:

aa a a a a

div e R i L

dt (1)

where ae K n : electromotive force,

e m aT K i : electromagnetic torque (Nm),

LT : Load torque (Nm), av : DC source

voltage (V), aR : Armature resistance (), aL :

Armature inductance (H), and ai : Armature

current (A).

The torque equivalent equation of DC motor:

e L

dJ T T

dt

(2)

where is the angular speed. and J is the

inertia constant.

Due to 2

60 9.55

n n so

9.55( )e L

dnT T

dt J (3)

We choose state variables: 1 2;ax i x n ,

we get

11 2

21

1

9.55 9.55

a ea

a a a

mL

R Kdxx x

dt L L L

Kdxx T

dt J J

(4)

In the other hand: 2e mK K , (4) becomes

11 2

21

2 1

9.55 9.55

a ma

a a a

mL

R Kdxx x

dt L L L

Kdxx T

dt J J

(5)

DIGITAL CONTROL OF A AC/DC CONVERTER

Three phase full wave rectifier

Three phase full-wave converter is a fully

controlled rectifier using six thyristors. All the six thyristors are controlled switches which are

turned on at a appropriate times by applying suitable gate trigger signals. The three phase

full wave converter is extensively used in industrial power applications up to about

120kW output power level, where two quadrant operations are required. The figure

shows a three phase full - wave converter with highly inductive load. The thyristors are

triggered at an interval of (π/3) radians (i.e. at an interval of 60). The frequency of output

ripple voltage is 6fs and the filtering requirement is less than that of three phase

semi and half wave converters.

At t = (π/6 +α), thyristor T6 is already

conducting when the thyristor T1 is turned on by applying the gating signal to the gate of T1.

During the time period t = (π/6 +α) to (π/2 +α), thyristors T1 and T6 conduct together and

the line to line supply voltage vab appears across the load.


89

M

a

R4C4

R6C6

R2C2 R5C5

R3C3

R1C1

T4

T6

T2 T5

T3

T1

b c

i1i1i4i4

i3i3i6i6

i5i5i2i2

vg1vg1

vg3vg3

vg5vg5

vg4vg4

vg6vg6

vg2vg2

Figure 1. The three phase full wave rectifier with load

At t = (π/2 +α), the thyristor T2 is triggered

and T6 is reverse biased immediately and T6

turns off due to natural commutation. During

the time period t = (π/2 +α) to (5π/6 +α),

thyristor T1 and T2 conduct together and the

line to line supply voltage vac appears across

the load.

The thyristors are numbered in the circuit

diagram corresponding to the order in which

they are triggered. The trigger sequence (firing

sequence) of the thyristors is 12, 23, 34, 45, 56,

61, 12, 23, and so on. The figure shows the

waveforms of three phase input supply

voltages, output voltage, the thyristor current

through T1 and T4, the supply current through

the line ‘a’.

Control card - TMS320F28069

The DSP processor used here is

TMS320F28069. This DSP processor has

speed up to 80 MHz and provides the power of

the C28x core and CLA coupled with highly

integratedcontrol peripherals in low – pin

count devices. It has enhanced peripherals such

as high resolution PWM module and 12 bit

A/D converter with conversion speed up to

12.5ns. It also has 32x32 bit multiplier, 32 bit

timers and real time code debugging capability

which gives all the benefits of the digital

control and also allows implementation of high

bandwidth.

Zero – crossing detector

To use PWM to send pulse to gates of

Thyristors, you need to synchronize the pulses

of PWM with Zero crossing point, so we need

to have Zero crossing detector.

Figure 2. The zero crossing detector circuit

Zero crossing detector is drawn on the above

figure. This circuit is supplied the AC power

from a isolation transformer via port P1. This

voltage then is rectified half - wave using

diode D1 and C1 to supply power to IC U1A

LM358. So this circuit is called zero crossing

detector.

EXPERIMENTAL RESULTS

Control system scheme

Figure 3. The control system scheme

The above structure describes the principle of

the whole system; to provide voltage to the

motor armature, we use the controlled three -

phase full wave rectifier. The purpose of this

system is when the load changes, the speed of

the motor keeps constant. To do this, the

negative feedback signal of the speed will be

taken back to compare with the reference, via

the controller PI controller, the system will

send the pulses to the Thyristors so that

keeping the speed constant.


90

Experimental setup

Figure 4. The experimental setup

- Motor – load system

- Feedback circuit

- Zero - crossing detector and 1 phase

synchronous transformer

- TMS320F28069 Board

- Amplifiers and DC transformer

In this system, we do not use pulse divider, but

in each time, we need to send two pulses to

two Thyristors to ensure that the current can

run continuously. To do above thing, we use

PWM pulse generator synchronizing with zero

crossing point from outside via GPIO pins.

Experimental results

In the following are the experimental results,

Figure 6 shows the waveform of the output

voltage, it looks very good. Each time, there

are two Thyristors working so we can see that

in one circle, there are six similar intervals of

the output voltage.

Figure 5. The waveforms of the output voltage

Fig. 6 shows the speed of the motor tracking

the reference very quick even when we change

the value of the reference. This proves that the

digital control system meets the requirements

about responding time and the tracking quality.

Figure 6. The speed of motor and its reference value

CONCLUSIONS

The speed control of DC motor drive using

TMS320F28069 digital signal processor is

achieveed. Due to TMS320F28069 has highly

efficient computation ability and rich operation

functions, so the PI controller is completely

achieved by software. The experimental results

show that the shape of armature voltage is

good and the speed track the reference vary

well. However, the current control loop in this

paper we still don’t finish, and here is the

future work for our job

REFERENCE

1. TS. Lưu Kim Thành, ThS. Bùi Tuấn Anh, “

Thực hiện kỹ thuật điều khiển số hệ truyền động

điện một chiều DC Servo với Card Advanced PCI-

1711 trên Matlab-Simulink”, Tạp chí Công nghệ

hàng hải, 9/2009.

2. PGS.TSKH. Nguyễn Phùng Quang, Điều khiển

số - Digital control, Bài giảng dành cho học viên

cao học ngành TĐH và ĐKTĐ trường ĐH Bách

Khoa Hà Nội, 2007.

3. Võ Như Tiến, Phan Điền, “ Ứng dụng xử lý tín

hiệu số DSP điều khiển tốc độ động cơ một

chiều”, Tạp chí Khoa học-Công nghệ, Đại học Đà

Nẵng, 2009

4. Chi-Tsong Chen, Analog and Digital control

system design, Sauders College Publishing.

5. Dogan Ibrahim, Microcontroller Based Applied

Digital Control, John Wiley & Sons, Ltd, 2006.

6. Taigiang Cao, Jianping Xu, and Shungang Xu, “

Designing DSP Based Digital Control DC Motor

System”, IEEE 2008.


91

TÓM TẮT ĐIỀU KHIỂN SỐ ĐỘNG CƠ MỘT CHIỀU

QUA MẠCH THYRISTOR CÔNG SUẤT

Vũ Ngọc Huy*, Trần Mạnh Tuấn, Nguyễn Thị Mai Hương, Nguyễn Tiến Hưng


Các động cơ DC đã được sử dụng rộng rãi trong công nghiệp vì công suất cơ lớn, đơn giản và chi

phí hợp lý cũng như các đặc tính điều khiển linh hoạt của chúng. Tốc độ của động cơ DC có thể

được điều khiển bằng cách điều chỉnh điện áp phần ứng sử dụng các mạch thyristor công suất.

Trong quá khứ, điều khiển tốc độ của động cơ DC đã được thực hiện với kỹ thuật tương tự do đơn

giản, và chi phí thấp. Tuy nhiên, kỹ thuật này thể hiện nhiều hạn chế vì nó không thể đảm bảo điều

khiển khi có sự thay đổi của các thông số thiết bị trong một phạm vi khi các điều kiện làm việc

thay đổi. Bài báo này đề xuất việc điều khiển tốc độ của động cơ DC sử dụng một bộ xử lý tín hiệu

số (DSP) mà có các tính năng đặc biệt cho điều khiển số động cơ. Vòng điều khiển tốc độ sử dụng

bộ điều khiển PI được thực hiện bằng cách lập trình trong DSP. Dựa theo tín hiệu vào, tín hiệu

phản hồi và thuật toán điều khiển, các xung PWM tạo ra bởi các DSP sẽ được gửi đến cực G của

các thyristors ở những thời điểm thích hợp.

Từ khóa: Động cơ DC, xử lý tín hiệu số, DSP, PWM.



92


93

ROBUST CONTROL OF DC MOTORS

Huy Ngoc Vu*, Tuan Manh Tran, Huong Thi Mai Nguyen, Hung Tien Nguyen University of Technology – TNU

ABSTRACT

DC motors are used in a wide variety of applications touching our daily lives, where they serve to

relieve us in many of works. However, there is still difficult in designing a control system for the

DC motors since the data used for designing controller do not match the measured data in reality.

This is because of the motor parameters are normally changing with time. Hence the controller

could not guarantee the perfect performance all the time in order to get a highest effectiveness. In

this research, the parameters of the motor are identified by using Matlab software. After that an

controller is designed in order to ensure that the performance of the controlled system is

maintained with respect to the changes of motor parameters in specified ranges. The research

results are presented by some simulation results in Matlab/Simulink environment.

Keywords: DC-motor, uncertain parameter, H-infinity, effectiveness.

INTRODUCTION*

It is well-known that the electrical parameters

of the machine are strongly affected by

temperature, magnetic saturation, and

winding current modulus. These effects can

deteriorate the controller performance when

designed with nominal parameter values.

Therefore, a better performance requirement

against changes in the machine parameters

and exogenous inputs is desired to be

achieved by appropriate controller design for

DC motors. In [1], the classical technique

control which is PI controller has been

studied for servo application to drive the

system or loads to control speed and torque.

The objectives of this work are to obtain good

performances of the DC drive which are

minimum of overshoot and less sensitive to

the parameters variations. Also, this controller

has a good dynamic response such as to

provide fast transient response and has been

applied. In this work, the torque and speed

control is been discussed. In the first stage,

the PI controller is designed to get the proper

value of PI controllers with Linear Theory

Analysis. Then the PI values of the torque and


speed will be evaluated under a large signal

analysis. In [2], to identify the parameters of

the motor, an experimental measurement of

armature voltage, armature current and rotor

speed are performed using the NIDAQ USB-

6008 data acquisition module. Trials are

performed to apply some of the methods on

the motor practically to identify the motor

parameters. DAQ toolbox in Matlab/Simulink

is used to acquire the test signals and perform

analysis based on the nonlinear least square

method or pattern search method which they

are suitable. Finally, we designed a GUI to be

user friendly and to automate all the process

of identification.

From the above overview, we realize that the

difficulties to control the motor are the

unknown and uncertain parameters; these

make the responses not good, so in this paper,

the authors will estimate the motor’s

parameters using Matlab software and then

apply H1H1 to control to get the robust

stability.

THE MATHEMATIC MODEL OF DC MOTOR

The voltage equivalent equation of DC motor:

aa a a a a

div e R i L

dt (1)


94

where ae K n : electromotive force,

e m aT K i : electromagnetic torque (Nm),

LT : Load torque (Nm), av : DC source

voltage (V), aR : Armature resistance (), aL :

Armature inductance (H), and ai : Armature

current (A).

The torque equivalent equation of DC motor:

e L

dJ T T

dt

(2)

where is the angular speed. and J is the

inertia constant.

Due to 2

60 9.55

n n

so 9.55

( )e L

dnT T

dt J (3)

We choose state variables: 1 2; ,ax i x n

we get

11 2

21

1

9.55 9.55

a ea

a a a

mL

R Kdxx x

dt L L L

Kdxx T

dt J J

(4)

In the other hand: 2e mK K , (4) becomes

11 2

21

2 1

9.55 9.55

a ma

a a a

mL

R Kdxx x

dt L L L

Kdxx T

dt J J

(5)

PRELIMINARIES IN ROBUST CONTROL

Let L2 denote the space of square-integrable

signals defined on the interval [0,∞). A matrix

A is called symmetric if it is real and satisfies

A=AT. The set A=AT of all mxm symmetric

matrices will be denoted by Sm.

The H∞-norm

Consider a linear input-output system

that is described by

:x Ax B

z Cx D

(6) (6)

and whose transfer matrix is given by

1( ) ( )G s C sI A B D

If AA is stable and if we choose the initial

condition x(0)x(0) to be zero, §§ defines a linear

map w ! zw ! z on L 2L 2 with a finite energy gain

defined as

2

2

, 02

supL

G

It is well-known that the energy-gain of

coincides with the H∞-norm of the

corresponding transfer matrix G given by

_

sup ( ( ))R

G G j

Where _

( )M stands for the largest singular

value of the complex matrix matrix M.

The bounded real lemma

It is not possible to explicitly compute G

in

terms of the realization matrices. Instead, one

can characterize stability of G and the validity

of the inequality

G (7)

as an LMI in some auxiliary matrix K and ,

which is one version of the celebrated

bounded real lemma. Indeed, it can be shown

[3] that A is stable and that (7) holds if and

only if DTD-2I<0 and there exits some

KT=K>0 such that [4]

2

0

0T T T

T T T

K

A K KA C C KB C D

B K D C D D I

This result is referred to as the bounded real

lemma.


95

H∞ control of linear time-invariant systems

A standard setup for H∞ control is presented

in Fig. 1, where represents the generalized

disturbances, z the controlled variable, u the

control input and y the measuement output,

while P is a linear time-invariant system

described as

p

p p

x Ax B Bu

z C x D Eu

y Cx F

PP

ww zz

KK

uu yy

Figure 1. The interconnection of the system

The goal in H∞ control is to find a stabilizing

LTI controller K that minimizes the H∞ norm

of the closed-loop system ( , )lF P K

, where

( , )lF P K is lower linear fractional

transformation of P and K, which is nothing

but the closed-loop transfer function z

in Fig. 1.

In order to achieve certain desired shapes of

the closed-loop transfer functions, such as

dictated by requirements on bandwidth,

weights are introduced and we consider

minimizing the H∞-norm of ( , )lF P K

,

where ( , )lF P K is the closed-loop transfer

function z in Fig. 2, Wz and W are

real-rational proper weighting functions with

suitable band-pass and characteristics.

Sub-optimal H∞ control

Let us now consider a generalized plant P

where weights are incorporated already as

follows

0

p

p p

x A B B x

z C D E

y C F u

(8)

PP

ww zz

KK

uu yy

WwWw WzWz

ezezewew

ePeP

Figure 2. The weighted interconnection of the system

If the linear time-invariant controller KK is

expressed as

c cc c

c c

A Bx x

C Du y

(9)

the closed-loop system ( , )lF P K admits the

following state-space description:

A B

C Dz

(10)

where

c c p c

c c c

p c c p c

A BD C BC B BD FA B

B C A B FC D

C ED EC D ED F

(11)

The H∞ control problem is to find an LTI

controller which renders A stable and such

that

A B

C D

(12)

holds true [5], where >0 is a given number

that specifies the performance level. This is

the so-called sub-optimal H∞ problem.


96

H∞ controller synthesis

Using the bounded real lemma for (12), the

matrix A is stable and (12) is satisfied if and

only if the LMI

holds for some X>0. Unfortunately this

inequality is not affine in X and in the

controller parameters which are appearing in

the description of A, B, C, D. However, a by

now standard procedure allows to eliminate

the controller parameters from these

conditions, which in turn leads to convex

constraints in the matrices X and Y that appear

in the partitioning of

according to that of A in (11). One then

arrives at the following synthesis LMIs for

H∞-design [6]:

(13)

where X and Y are basis matrices for the

subspaces

respectively.

After having obtained X and Y that satisfy

(13) for some level , the controller

parameters can be reconstructed by using the

projection lemma [7]. This procedure for H∞-

synthesis is implemented in the robust control

toolbox [8].

Mixed sensitivity H∞ approach

A standard setup for H∞ control as depicted in

Fig. 3. For this control configuration,

engineers are usually interested in some

specific transfer functions. In particular, S =

(I + GK)-1 is the sensitivity function which

describes the influence of the external

disturbance ww to the tracking error yy. T = (I +

GK)-1GK is the complementary sensitivity

function which describes the influence of the

reference signal to the system output z.

Finally, KS is the transfer function from to

the control input u that indicates control

activity.

GG

uu

KK

yy

zzw

++

¡¡

PP

Figure 3. Feedback control configuration

Mixed sensitivity H∞ performance

In general, performance of the closed-loop

system that is specified by H∞ norm of the

channel z can be formulated as a multi-

objective problem (see Fig. 4). This leads to

the minimization of

(14)


97

The multi-variable loop shaping with various

specifications (14) is the so-called the mixed

sensitivity H∞ design approach.

GGuu

KKyyw

++ ¡¡

z1z1

z2z2

z3z3

zz

Figure 4. Multi-objective problem

It is well-known in the literature that the

transfer functions S, T, and KS need to be

small in magnitude in order to achieve good

command tracking and robust stability.

However, the well-known constraint S+T=I

reveals that these requirements can not be

achieved simultaneously over the whole

frequency range. However, the use of

frequency filters or weighting functions opens

up the possibility to minimize the magnitudes

of S, T, and + over different frequency ranges.

Hence, in practice, instead of minimizing (14)

one rather determines a stabilizing LTI

controller K that minimizes the cost function

(15)

here WS, WP, WT are suitably chosen

weighting functions (Fig. 5).

GGuu

KKyyw

WTWT

WPWP

WSWS

++ ¡¡

zz

Figure 5. Mixed sensitivity H∞ control

DESIGNING AN H∞ CONTROLLER FOR

A DC MOTOR

Due to Ra, La, Km are uncertain parameters,

we can represent them as follow

0

0

0

(1 )

(1 )

(1 )

a a r r

a a l l

m m m m

R R

L L

K K

where Ra0, La0, Km0 are the nominal

parameters; r, l, m and -1 r, l, m 1

represent the variations of the system

parameters.

The model of the DC motor with uncertainties

is shown in Fig. 6.

RRvava

RR

- -

+

-

+

_x1_x1

_x2_x2

x1x1

x2x2

©©

2¼©2¼©

M rM r

±r±r

M lM l

±l±l

MmMm

±m±m

MmMm

±m±m

TLTL

urur

yryr

ununynyn

uiuiyiyi

ulul ylyl

vivi

vrvr

vnvn

9.55

J

Figure 6. The uncertain model

This model can be rearrange to the M ¢M ¢

configuration with matrix MM given by


98

0 1 0 1

0 0 0 0 01

2 2 0 2 2

0 1 0 1

0 0 0 0 0

0

0

01

2

10 0

0 0 0 0 0

10 0

0 0 0 0 0 0 0

0 0 0 0 0 0 0

0 0 0 0 0 0 0

1 0 0 0 0 0 0 0

0 1 0 0 0 0 0 0

a m mrl

a a a a a

m m

l a m mrl

a a a a ar

an

mi

m

R K

L L L L Lx

x K

y R K

L L L L Ly

Ry

Ky

Ky

y

1

2

l

r

n

i

L

a

x

x

u

u

u

u

T

v

and the matrix is given by

i

n

r

l

m

m

r

l

i

n

r

l

y

y

y

y

u

u

u

u

000

000

000

000

=

The certain model of the DC motor can be

represented as in Fig. 7.

GmGm

ulul

urur

unun

vava

TLTL

ylyl

yryr

ynyn

yiyi

y1y1

uiui

y2y2

Figure 7. Certain model of the DC motor

SIMULATION RESULTS

Frequency response

100

101

102

103

104

105

106

-150

-100

-50

0

From: nref To: [+Gupss]

Magnitude (

dB

)

Closed-loop performance of reference inputs to outputs

Frequency (rad/s)

Figure 8. Output response with reference input

(speed)

100

101

102

103

104

105

106

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

5

From: nref To: [+nref-Gupss]

Magnitude (

dB

)

Closed-loop performance of the load to outputs

Frequency (rad/s)

Figure 9. Reference input with error

100

101

102

103

104

105

106

-140

-130

-120

-110

-100

-90

-80

-70

-60

-50

From: TL To: [+Gupss]

Mag

nitu

de (

dB)

Closed-loop performance of reference inputs to outputs

Frequency (rad/s)

Figure 10. Effect of the load TL to output speed

100

101

102

103

104

105

-120

-100

-80

-60

-40

-20

0

20

40

From: nref To: [+Gupss]

Magnitude (

dB

)

Bode Diagram

Frequency (rad/s)

Figure 11. Output Response

100

101

102

103

104

-50

-40

-30

-20

-10

0

10

20

30

40

50

From: nref To: [+nref-Gupss]

Magnitu

de (

dB

)

Bode Diagram

Frequency (rad/s)

Figure 12. Error


99

Time response

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.050

0.2

0.4

0.6

0.8

1

1.2

1.4Closed-loop performance of reference inputs to outputs

time (s)

Figure 13. Step response

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05-0.2

0

0.2

0.4

0.6

0.8

1

1.2Closed-loop performance of reference inputs to control errors

time (s) Figure 14. Error

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0x 10

-3 The effect of torque to output

time (s)

Figure 15. The effect of torque to output

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.050

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8x 10

-3 The effect of torque to control errors

time (s)

Figure 16. The effect of torque to control errors

CONCLUSIONS

The paper presents an approach for the robust

controller design for the DC motor. The

approach is based on uncertain model of the

motor. Consequently the H1H1 controller was

designed via Matlab functions. The controller

is of the second order, thus quite simple. It

was tested in a simulation and it was proved

that the controller is able to stabilize the

system even the most degraded model within

the given uncertainty range. Such a controller

should be theoretically able to stabilize also

the real system with behavior covered by the

uncertainty. The approach is possibly

applicable to other systems where it is

impossible to create precise model for the

control design.

REFERENCES

1. Dayang. F. B, “Control of DC motor using PI

controller”, Thesis, 2005.

2. Mohammed S.Z, “Parameters identification of

a permanent magnet DC motor”, Master thesis,

2009.

3. K. Zhou and P. P. Khargonekar. An algebraic

riccati equation approach to H1H1 optimization.

Systems and Control Letters, 11:85–91, 1988.

4. J. G. Van Antwerp and R. D. Braatz. A tutorial

on linear and bilinear matrix inequalities. Journal

of Process Control, 10:363–385, 2000.

5. P. Apkarian and P. Gahinet. A convex

characterization of gain-scheduled H1H1

controllers. IEEE Transactions on Automatic

Control, 40:853–864, 1995

6. C. W. Scherer. Mixed H2/H1H1 control for time-

varying and linear parametrically-varying systems.

International Journal of Robust and Non-linear

Control, 6:929 – 952, 1996.

7. C. W. Scherer and S.Weiland. Linear Matrix

Inequalities in Control. Lecture notes in DISC

course, 2005.

8. A. Packard M. Safonov G. Balas, R. Chiang.

Robust control toolbox for use with Matlab,

volume 3. The MathWorks, 2005.


100

TÓM TẮT

ĐIỀU KHIỂN BỀN VỮNG ĐỘNG CƠ MỘT CHIỀU

Vũ Ngọc Huy*, Trần Mạnh Tuấn, Nguyễn Thị Mai Hương, Nguyễn Tiến Hưng


Các động cơ DC được sử dụng trong rất nhiều ứng dụng trong cuộc sống, giải phóng sức lao động

của con người trong nhiều công việc. Tuy nhiên, vẫn còn khó khăn trong việc thiết kế một hệ

thống điều khiển cho động cơ DC bởi vì dữ liệu được sử dụng để thiết kế bộ điều khiển không phù

hợp với dữ liệu đo được trong thực tế. Điều này là do các thông số động cơ thường thay đổi theo

thời gian. Do đó bộ điều khiển không thể bảo đảm thực hiện hoàn hảo để có hiệu suất cao nhất.

Trong nghiên cứu này, các thông số của động cơ được xác định bằng cách sử dụng phần mềm

Matlab. Sau đó một bộ điều khiển H∞ được thiết kế để đảm bảo rằng hiệu suất của hệ thống được

điều khiển được duy trì với những thay đổi của các thông số động cơ trong một dải xác định. Các

kết quả nghiên cứu được đưa ra bởi các kết quả mô phỏng trong môi trường Matlab / Simulink.

Từ khóa: Động cơ DC, tham số bất định, H-vô cùng, hiệu suất.


Kien Ngoc Vu et al Journal of SCIENCE and TECHNOLOGY 127(13): 101 - 106

101

MODEL REDUCTION IN SCHUR BASIS WITH POLE RETENTION

Kien Ngoc Vu1,*, Du Huy Dao1, Cong Huu Nguyen2

1University of Technology – TNU; 2Thai Nguyen University ABSTRACT

Model order reduction is a research direction which is more interested scientists in recent years.

There have been many order reduction algorithm introduced to many different approaches in

which retaining the important poles of the original system in the reduced root system is the right

approach and has many advantages.

This paper presents a new model order reduction algorithm, the order reduction algorithm based on

Schur analysis, based on the idea of keeping the important poles of the original system in the order

reduction process. This algorithm transforms matrix A of the higher-order original system to upper

- triangle matrix on which the poles are arranged in descending important properties on the main

diagonal of the upper – triangle matrix. The illustrative examples show the correctness of the

model order algorithm.

Keywordss: Model order reduction, Schur analysis, important poles.

INTRODUCTION*

In the previous paper [1] the authors

introduce balanced truncation algorithms. The truncation [1] for the system is based on

Hankel singular value (It removes the state corresponding to Hankel small singular

values) leading to important climax points of the root system without conserved reduced

order system. However, important climax poles (dominant poles) are invariant in the

real system, so it should be preserved in process of order reduction. Therefore, this

paper, we introduce a new algorithm, the order reduction algorithm based on Schur

analysis, based on the idea of keeping the important poles of the original system in the

order reduction process. The illustrative examples shows the correctness of the model

order algorithm.

MODEL ORDER REDUCTION ALGORITHM

Problem of order reduction model

A linear system is given with continuous-time

constant parameters available multiple-inputs multiple-outputs described in state space by

the following equations:

x=Ax+Bu

y=Cx(1)


In which, x Rn, u Rp, y Rq, A Rnxn, B

Rnxp, C Rqxn.

The goal of the order reduction problem with

model described by (1) is to find models

described by systems of equations:

r r r r

r r r

x =A x +B u

y =C x(2)

In which, xr Rr, u Rp, yrRq, Ar Rrxr, Br

Rrxp, Cr Rqxr, với r n;

So that the model described by (2) can be

replaced by the model described in (1) to

apply in analysis, design and control system.

Model reduction in Schur basis with pole

retention

Model reduction in Schur basis with pole

retention was developed by the researched

team based on truncation technique and

analysis of Schur.

Truncation technique is a method of simple

order reduction. In which, the main idea of it

can be divided into 2 steps: the step 1 is to

convert the high original system to the

equivalent system with a non-singular

transformation in state space. The step 2 is to

delete rows and columns of similar systems to

generate the reduced order system. The two

most typical algorithms for truncation


102

technique are balanced truncation [1] and

model truncation [2] in many disorders.

However, disadvantages of both methods are

the use of the Singular Value Decomposition

(SVD) which required a lot of bidding

conditions. On the other hand, the truncation

for the system is based on Hankel singular

value (It removes the state corresponding to

Hankel small singular values) leading to

important climax points of the root system

without conserved reduced order system.

Important climax poles (dominant poles) are

invariant in the real system, so it should be

preserved in process of order reduction.

From those disadvantages, the key idea of

algorithm provided by the author, Minh H.B

and his research group [3] is to convert the

matrix A of the internal system to (1) of the

triangular matrix based on Schur analysis

(without SVD analysis), on the basis the poles

are arranged in descending important

properties on the main diagonal of the upper

triangular matrix A. Then go to the step 2 of

the truncation technique, which helps

important climax poles (dominant poles) to be

preserved in the reduced order system.

Order reduction algorithm is as following

Algorithm 2.2.1 (Triangle realization)

Assume that the linear time-invariant system

(A, B, C, D) is asymptotically stable and in a

minimal representation.

Input: The original system (A, B, C, D)

Step 1: Compute observability Grammians

Q from Lyapunov equation A*Q + QA +C*C = 0

Step 2: Compute Cholesky factorization

Q = R*R.

Step 3: Compute Schur decomposition of

RAR-1:

RAR-1=UU*, where U is unitary matrix and

is upper triangle matrix.

Step 4: Compute non-singular

transformation T = R−1U

Step 5: Compute A, B, C, D = (T-

1AT, T-1B, CT,D)

Output: An equivalent system with realization

A, B, C, D

The output realization in Algorithm 2.2.1 is

said to be triangle realization.

Algorithm 2.2.2 (Re-ordering the poles by

dominance index)

Input: Triangle realization A, B, C, D ,

which is the output of Algorithm 2.2.1

Step 1: For each pole i, i = 1 … n, compute

its dominance index i i

2

i

i

C BR =

Reλ.

Step 2: Choose the largest dominance index 1i

R

Step 3: Reorder pole 1i

λ (and its conjugate

1iλ , if it appears) to the first position in the

diagonal of A by unitary matrix U1:

Comment of Step 3: Algorithms for

reordering eigenvalues in Schur

decomposition are referred to [7, 8]. It can be

done, for example, by MATLAB function

ordschur

Step 4: Compute new equivalent realization

(* *

1 1 1 1U AU , U B, CU ).

Step 5: Remove two first rows and columns

of (* *

1 1 1 1U AU , U B, CU ) to obtain a smaller

realization ( ˆ ˆÂ, B, C ) with (n − 2)

dimensions.

Step 6: Do the same procedure from Step 1 to

Step 5 for smaller realization ( ˆ ˆÂ, B, C ) and

continue this loop until all poles are re-

ordered.


103

THE ILLUSTRATIVE EXAMPLES

Reducing higher-order controller

H full order controller of the robot balanced

system [4] has the following transfer function

model: 6 5 4

3 2

8 7 6 5 4 4

3 2 5

0.1209 0.01874 27.07

2.077 1505 6.713 2853( )

0.3542 390 50.7 3.53

936.7 4.674 786.2 1.472

s s s

s s sW s

s s s s e s

s s s e

(3)

The full order controller 6th-order will lead to

many disadvantages when we perform robot

balanced control programs due to complex

program that make increase of processing

time, the slow speed of response of the

control system, without a good response to

real-time requirements of the controller and

instability of balanced system. So, in order to

improve the quality of this controller, it

should implement reduced order controller to

let program code be simpler, reduction of

processing time, increase of response speed,

but still satisfy requirements of sustainable

stability of the system.

Implementation of reduced order H control

algorithm reduced sufficient steps mentioned

above; we obtain the following result table:

Table 1. Results of the order reduction controller

of the robot balanced system

Order Transfer function – Wcr(s)

5

4 3 2

5 4 3 2

1275 8.694 5 4.367 5

1.359 8 1.209 7

715.6 2.349 4 2.768 5

3.777 6 3.183 5

s e s e s

e s e

s s e s e s

e s e

4 9.46655409.39787.33

4773.15993.11.3481275234

23

ssss

esess

3 550639578.33

5993.18.234127523

2

sss

ess

2 43.9425.30

6.24711302

ss

s

1 71.26

1006

s

Evaluation of the reduced order error based

on H norm, we obtain the following result

table.

Table 2. Error between the 6th-order controller

and reduced order controller

Order Error

Hcr ss )(W)(Wc

5 5.1995e-005

4 4.3560e-004

3 1.7910

2 37.2364

1 38.1419

To verify the reduced order model, the

researchers have simulated the excessive

response of the full order controller and

reduced order controllers. The simulation

results in Matlab/Simulink as shown in Fig 1.

Figure 1. Results of simulation of full order

controller and reduced order controller

To compare the effectiveness of the new

order model reduction algorithm with the

algorithm based on Schur analysis of M. G.

Safonov and R. Y. Chiang [5], the authors

perform reduced order controller (3) using

algorithm in [5], we obtain the results in the

following table 3.



table 4.


104

Table 3. Results of the reduced order controller

using algorithm in [5]


5

3 2

5 4 2

4

3

.694 005 4.366 005

1.359 008 1.209 007

715.6 2.349 004 2.768 005

3.777 006 3.183 005

1275 8 e s e s

e s e

s s e s e s

e s

s

e

4 3 2

4 3 2

1275 348.2 1.993 005 1.775 004

33.87 397.9 5540 467.2

s s e s e

s s s s

3 2

3 2

1275 233.8 1.992 005

33.76 395 5499

s s e

s s s

2 2

946.3 s + 227.9

24.55 386s s

1 946.4

s + 24.55

Table 4. Error between the 6th-order controller

and reduced order controllers

Order Error

Hcr ss )(W)(Wc

5 1.5977e-006

4 3.4723e-004

3 1.7678

2 37.3971

1 38.8756

From the result of order reduction according

to two methods we see: Compared to

response h(t) of the full 6th-order controller,

the response h(t) of the 5th, 4th-order reduction

controller coincide completely; response h(t)

of the 3rd-order reduction controller has very

small differences; response h(t) of 2nd and 1st-

order reduction controllers, has greater

differences. Therefore we can use the 5th, 4th

and 3rd-order reduction controller to replace

the full 6th orders controller.

Comparison of the poles between 6th-order

controller and the 3rd-order reduction

controller, we obtain the result as follows:

+ Poles of 6th-order original controller

-681.74; -26.71; -3.5353 + 13.9156i; -3.5353;

- 13.9156i; -0.09; -0.08

+ Poles of 3rd-order reduction controller using

algorithm in [5]: -26.6843; -3.5396 +13.912i;

-3.5396 -13.912i

+ Poles of 3rd-order reduction using new

algorithm: -26.71; -3.5353 +13.9156i; -3.5353

-13.9156i

The important pole of the original 6th-order

controller has been preserved in the 3rd-order

reduction system using new algorithm, with

algorithm of M. G. Safonov and R. Y. Chiang

[5], the important pole of the original 6th-

order controller is not conserved in the 3rd-

order reduction system.

Reducing 8th–order SISO system

A 8th–order SISO system often is selected to

evaluate and compare between the model

order reduction algorithm is given in [6] as

follows:

( )( )

( )

N sW s

D s (5)

with: N(s) = 0.1209s6 + 0.01874s5 + 27.07s4 +

2.077s3 + 1505s2 + 6.713s + 2853

D(s) = s8 + 0.3542s7 + 390s6 + 50.7s5 +

5.53e4s4 + 936.7s3 + 4.674s2 + 786.2s +

1.472e5

Simplification of the model system will help

the process of understanding the system

easier. The authors perform reduced order

model system based on Schur analysis, the

results shown as follows:

Table 5. Results of the order reduction of the 8th-

order SISO system


7

6 5 4

3 2

7 6 5 4

3 2

0.0001106 0.1184 0.05018

13.12 0.6966 25.44 0.4283

0.3436 278.1 10.46

4162 50.89 1314 13.93

s s s

s s s

s s s s

s s s

6

5 5 4 3

2

6 5 4 3

2

1.011 0.1185 0.01536

13.12 0.05516 25.44

0.333 278 7.516

4162 6.772 1314

e s s s

s s

s s s s

s s

5 4 3 2

5 4 3 2

0.004671 0.1116 1.317 11.57 19.14

0.3324 277.7 7.256 4072 2.24

s s s s

s s s s s


105


4

6 3 2 5

4 3 2

5.755 0.04525 9.796 0.09565

0.009 15.85 0.01963 5.012

e s s e s

s s s s

3

2

3 2

0.004693 0.03988 0.07292

0.00845 15.53 0.008542

s s

s s s

2

6

2

6.438 0.005329

0.0011 0.3227

e s

s s

1 0.004694

0.00055s



table 6.

Table 6. Error between the 8th-order SISO system

and reduced order systems

Order Error ( ) ( )rW s W s

7 0.0114

6 0.0109

5 8.5286

4 0.0140

3 8.5199

2 1.2822

1 8.5199

Bode plots of the original system, reducted

order systems are shown Fig.2

Figure 2. Bode plots of the original system,

reduced order systems

From the result of order reduction, evaluation of the reduced order error and bode plots, we see: the error between the original 8th-order SISO system and the 7th, 6th and 4th-order reduction system is very small; the error

between the original 8th-order SISO system and the 5th, 3rd, 2nd and 1tst-order reduction system is very big. Therefore we can use the 7th, 6th and 4th-order reduction system to replace the 8th-order SISO system.

The authors perform order reduction of 8th-

order SISO system using the algorithm of M.

G. Safonov and R. Y. Chiang [5], we obtain

the results in the following table:

Table 7. Results of the order reduction of the 8th-

order SISO system

Oder Tranfer function Wr(s)

7

6 5 4

3 2

7 6 5 4

3 2

0.0001204 0.1182 0.05242

13.11 0.7416 25.44 0.4662

0.3432 278 10.75

4162 55.5 1314 15.39

s s s

s s s

s s s s

s s s

6

5 4 3

2

6 5 4 3

2

1.15 005 0.1185 0.01582

13.12 0.0555 25.43

0.3332 277.9 7.518

4160 6.77 1313

e s s s

s s

s s s s

s s

5

4 3

2

5 4 3

2

0.002274 0.04519

0.04324 0.09553 0.02669

0.1695 15.86

2.564 5.015 0.8044

s s

s s

s s s

s s

4

3 2

4 3 2

1.148 005 0.04525

0.0001099 0.09565

0.009 15.85

0.01963 5.012

e s s

s

s s s

s

3 2

3 2

0.005077 0.005312 0.001657

0.005046 0.3227 0.001275

s s

s s s

2 2

1.289e-005 s + 0.0053

0.0011 0.3227

29

s s

1 0.004697

s + 0.0005502



table 8.

From the result of order reduction, evaluation

of the reduced order error, we see: the error

between the original 8th-order SISO system

and the 7th, 5th, 6th and 4th-order reduction


106

system is very small; the error between the

original 8th-order SISO system and the 3rd, 2nd

and 1tst-order reduction system is very big.

Therefore we can use the 7th, 6th, 5th and 4th-

order reduction system to replace the 8th-order

SISO system.

Table 8. Error between the 8th-order SISO system

and reduced order system

Order Error ( ) ( )rW s W s

7 0.0109

6 0.0109

5 0.0138

4 0.0140

3 1.2809

2 1.2822

1 8.5199

Thus reduced order results according to model order reduction algorithm based on analysis Schur are complete correct.

CONCLUSIONS

This paper introduces detailed order model reduction algorithms based on Schur analysis. The most important new feature of the algorithm is the ability to arrange the poles in the importance of decreasing on the main diagonal of the upper triangular matrix A and the ability to retain the important pole of the original model in order reduction model. The illustrative examples show the correctness of the model order algorithm.

REFERENCES

Cong N.H , Kien V.N , Du D.H , Research reduce

order model algorithm by balancing method,

Journal of Science and Technology of the Technical

University, Vol 80, pp. 34-39, 2011

A.C Antoulas, Approximation of Large – Scale

Dynamical Systems, SIAM, 2005.

Minh H.B and Kiyotsuga Takaba.(2011) “Model

reduction in Schur basic with pole retention and H

- norm error bound,” In: Proceedings of

international workshop on Modeling, Systems, and

Conrol 2011

Thanh B.T, and Manukid Parnichkun. (2008)

“Balancing control of Bicyrobo by particle swarm

optimization – based structure-specified mixed

H2/H control,” International Journal of Advanced

Robotic Systems 2008; 5(4): 395- 402.

M. G. Safonov and R. Y. Chiang, "A Schur Method

for Balanced Model Reduction," IEEE Trans. on

Automat. Contr., vol. 34, no. 7, July 1989, pp. 729-

733.

J. Rommes, Methods for eigenvalue problems with

applications in model order reduction, PhD thesis,

Utrecht University, 2007

Zhaojun Bai, J.W. Demmel, On swapping diagonal

blocks in real Schur form, Linear Algebra and its

Applications, Vol. 186, pp. 755, 1993.

D. Kressner, Block algorithms for reordering

standard and generalized Schur forms, ACM

Transactions on Mathematical Software, Vol. 32,

No. 4, pp. 521-532, 2006.

TÓM TẮT

THUẬT TOÁN GIẢM BẬC BẢO TOÀN ĐIỂM CỰC

DỰA TRÊN PHÂN TÍCH SCHUR Vũ Ngọc Kiên1,*, Đào Huy Du1, Nguyễn Hữu Công2

1Trường Đại học Kỹ thuật Công nghiệp – ĐH Thái Nguyên; 2Đại học Thái Nguyên

Giảm bậc mô hình là một hướng nghiên cứu đang được nhiều nhà khoa học quan tâm trong những

năm gần. Đã có rất nhiều thuật toán giảm bậc được giới thiệu với nhiều hướng tiếp cận khác nhau

trong đó bảo lưu các điểm cực quan trọng của hệ gốc trong hệ giảm bậc là một hướng tiếp cận đúng

đắn và có nhiều ưu điểm.

Bài báo giới thiệu một thuật toán giảm bậc mô hình mới, thuật toán giảm bậc dựa theo phân tích

Schur, dựa trên ý tưởng bảo lưu các điểm cực quan trọng của hệ gốc trong quá trình giảm bậc. Thuật

toán mới chuyển đổi ma trận A của hệ gốc bậc cao về dạng ma trận tam giác trên trong đó các điểm

cực được sắp xếp theo tính chất quan trọng giảm dần trên đường chéo chính của ma trận tam giác

trên. Các ví dụ minh họa thể hiện tính đúng đắn của thuật toán giảm bậc

Từ khoá: Giảm bậc mô hình, thuật toán Schur, điểm cực quan trọng.


Viet Quoc Vu Journal of SCIENCE and TECHNOLOGY 127(13): 107 - 110

107

IMPROVING THE EFFICIENCY OF CONVENTIONAL DRINKING-WATER-

TREATMENT PROCESSES IN THE REMOVAL OF ARSENIC

Viet Quoc Vu*


ABSTRACT Since Vietnam is one of the most rice producing countries in the world, a huge amount of rice hull

waste produced every year has been raising environmentally significant concerns. This study aims

to build a model of improving the efficiency of the treatment process of removing arsenic from

drinking water using activated carbon derived from rice hull. In this model, the efficiencyof

treatment process is optimized due to the combination of the advantages of iron and activated

carbon. The iron hydroxide phases can improve maximum adsorption capacity and the activated

carbon can offer a high surface area for adsorption.

Keywordss: Drinking-water-treatment, removing arsenic, rice hull, activated carbon.

INTRODUCTION

Arsenic contamination has been a

considerable concern in many areas in

Vietnam. Arsenic element existing in nature

as As(III) and As(V). It is important to

remove both species of arsenic from drinking

water. Many studies have been conducted to

produce activated carbon from rice hulls for

arsenic removal. However, the arsenic

removal rate of activated carbon is not high

enough to fulfill health standards.

The most common process used to produce

activated carbon is chemical process. During

this process, carbonization and activation

occur at the same time. A chemical

(dehydrating) agent such as zinc chloride is

used to decompose the cellulose of rice hull.

An issue during this process is thatthe

efficiency of the removing metal irons is still

limited. This study aims to improve

performance of activated carbon, in order to

improve the efficiency of the removing

arsenic irons from drinking water.

2A model of improving the efficiency of the

removing arsenic from drinking water

process

This process combines the advantages of

both, iron and activated carbon. The Iron

hydroxide phases increase the adsorption

capacity. Activated carbon can offer a high

surface area for adsorption.

Figure 1. Iron-impregnated activated carbon process for removing Arsenic irons*



108

Table 1. The description of producing Iron-impregnated activated carbon for removing Arsenic irons

Description Temperature Time Equipment

Step 1 Washing 100g Rice Hull RC Deionized water

Step 2 Drying 80°C 5 hours Furnace

Step 3 Mixing with 1.0 M NaOH solution RC 10 hours 40g NaOH,

1 L Deionized water

Step 4 Washing until the base is undetected in

the filtrate RC Deionized water

Step 5 Drying until constant weight 80°C Furnace

Step 6 Grinding and sieving RC Siever and Grinder

Step 7

Covering with aluminum foil, placing

in alumina crucible, capping with an

alumina cover

RC aluminum foil, alumina

crucible with lid

Step 8 Placing in furnace 800°C 2 hours Furnace

Step 9 Cooling RC

Step 10 Rinsing three times in deionized water RC Deionized water

Step 11 Drying Furnace 110°C 10 hours Furnace

Step 12 Cooling RC

Step 13

FeCl2 (5.5g) is dissolved in 100ml

deionized water. Adding NaOH to

ferrous chloride solution until pH 12

RC

5.5g FeCl2,

100 ml Deionized

water, NaOH

Step 14 Adding AC (30g) to solution without

headspace RC

Glass bottle with lid,

250ml

Step 15 Shaking 24 hours Rotator

Step 16 Filtered Filter Bag

Step 17 Drying 110°C 10 hours Furnace

Step 18 Cooling RC

Step 19 Mix Fe-AC with 1M NaOH for 24

hours RC 24 hours

40g NaOH

1L Deionized water

Step 20 Soaking RC 24 hours 100 ml 38% HCl

Step 21 Washing and Drying RC Deionized water

RC: Room temperature


109

DISCUSSION

Choosing chemical agent NaOH

In this process,the silica in the rice hulls

reacts with NaOH forming sodium silicate

which leads to a porous surface area. Using

NaOH as a chemical agent results in a

comparable low surface area with a big pore

volume. It is used here for two reasons.

Firstly,Qiganget al. [1] state in their report

that the best results of coating activated

carbon with iron take place using a macro

pore activated carbon which means a low

surface area and a large pore volume. This is

because the large pores Fe3+ can enter deep

into the activated carbon. Using activated

carbon with small pores leads to an

impregnation only on the surface of activated

carbon. Secondly using NaOH will simplify

the process since it is also used for further

steps. The impact of NaOH on rice hulls is

shown in the picture stated below.

Choosing impregnation with Fe2+

In this process impregnation with Fe2+ is

chosen. Iron forms an amorphous layer of

iron oxides on the activated carbon surface.

Due to this layer the net positive surface

charge of the activated carbon is increased

and therefore the arsenic removal capacity of

activated carbon is enhanced. Arsenic

adsorption to iron oxide-hydroxide surfaces

can be described by the ligand exchange

mechanism. In the adsorption process,

arsenic species can replace hydroxyl ion

(OH-) on the surfaces of iron oxide-

hydroxides, forming inner-sphere complexes

[3]. Furthermore ferrous is soluble at a wide

range of pH and can diffuse deep into the

internal pores of the activated carbon. The

degree of impregnation onto the activated

carbon surface is maximal when the net

charge of the activated carbon surface

becomes negative. The net surface charge is

highly dependent on the pH. Therefore the pH

of the ferrous solution is increased to pH 12

by adding NaOH.

CONCLUSION

The removal rates of toxic metals have been

very promising and with adaptions like

iron-impregnation the results can be even

improved. The model of the process to

produce iron-impregnated activated carbon

for arsenic removal is a combination of a

variety of processes and could improve

the arsenic removal rate of activated carbon,

in which iron hydroxide phases increase the

adsorption capacity and activated carbon offer

a high surface area for adsorption.

Figure 2. Impact of NaOH on rice hulls; (a) Raw rice hulls and (b) Rice hull treated with NaOH

(Adapted from [2])


110

REFERENCES

1. Qigang Chang et al. (2010, June). The

preparation of iron-impregnated granular activated

carbon for arsenic removal from water. Journal of

Hazardous Materials 184 2010, 515-522.

2. Mohd F. Taha et al. (2011). Removal of Ni(II),

Zn(II) and Pb(II) ions from Single Metal Aqueous

Solution using Activated Carbon Prepared from

Rice Husk. World Academy of Science,

Engineering and Technology 60 2011, 291-296.

3. HyonChong, Kim et al. (2009). Arsenic

removal from water using iron-impregnated

granularactivated carbon in presence of

bacteria. Journal of Environmental Science and

Health, Vol. 40 2010, 177-182.

4. Http://www.epa.gov/nrmrl/wswrd/dw/arsenic/

pubs/FundamentalsofAdsorptionforArsenicRemov

alfrom Water.pdf, 17.12.2013.

TÓM TẮT

NÂNG CAO HIỆU QUẢ QUÁ TRÌNH LOẠI BỎ THẠCH TÍN

TRONG NƯỚC SINH HOẠT

Vũ Quốc Việt*


Việt Nam là một trong những nước sản xuất gạo nhiều nhất trên thế giới, vì vậy một lượng lớn vỏ

trấu được tạo ra trong quá trình sản xuất gạo có thể gây ảnh hưởng đến các vấn đề về môi trường.

Việc tận dụng nguồn vỏ trấu cho các ứng dụng hữu ích đang là vấn đề được quan tâm lớn. Mục

tiêu của bài báo này là xây dựng một mô hình nâng cao hiệu quả của quá trình loại bỏ thạch tín

trong nước sinh hoạt bằng than hoạt tính được tạo thành từ vỏ trấu. Trong nghiên cứu này, hiệu

quả tối ưu của quá trình xử lý đạt được nhờ sự kết hợp ion sắt và than hoạt tính. Ion sắt nâng cao

tối đa hiệu quả hấp thụ thạch tín, trong khí đó than hoạt tính được hình thành với lượng lớn diện

tích bề mặt hấp thụ.

Từ khóa: Xử lý nước uống, loại bỏ thạch tín, vỏ trấu, than hoạt tính.


Huyen Vu Xuan Dang et al Journal of SCIENCE and TECHNOLOGY 127(13): 111 - 115

111

ASSESSMENT OF TREATED LATEX WASTEWATER REUSE FOR

PERENNIAL TREE IRRIGATION ON GROUND WATER QUALITY

Huyen Vu Xuan Dang1,*, Hanh Vu Bich Dang1,

Amira Abdelrasoul2, Huu Doan2, Dan Phuoc Nguyen1 1Hochiminh city University of Technology, VNU-HCM, Vietnam

2Ryerson University, Toronto, Canada

ABSTRACT

The study aimed to assess nutrient contamination to aquifer from reuse of latex wastewater for

perennial tree irrigation. The latex wastewater contains high nitrogen concentrations and BOD that

are required high treatment costs to meet Vietnam Industrial effluent standards. The reuse of

secondary treated effluent for rubber tree irrigation may be a potential benefit in terms of treatment

cost reduction as well as nutrient reuse.

A pilot experiment was done in two lots of 100 rubber trees each. The area of each lot was 7m x

2.5m. One lot was controlled without irrigation. The used irrigation water was taken from effluent

of a facultative waste stabilization pond of latex processing industry. Hydraulic rate, nutrient

loading and COD loading applied to the lot were 8m3/ha/week, 12.5 kgN/ha/day and 0.5 kg COD

/ha/day respectively.

GMS modeling was used for assessing nitrogen and COD transport in the aquifer. The first order

reaction modeling was used for biological conversion of COD and nitrogen during the infiltration

in the vadose layer. The results shown that transport of total nitrogen and COD in the ground water

were reached stable levels at 2 meter depth after 10 months and 10 meter after 12 months of

irrigation. In addition, scanning electron microscopy (SEM) images used to compare the soil

quality among the sites.

Keywords: Latex processing wastewater, nutrient reuse, Vietnam Industrial effluent standards,

Groundwater Modeling System (GMS), scanning electron microscopy (SEM).

INTRODUCTION*

Vietnam Rubber Group reported that the

rubber industry emits by 10 million cubic

meters of wastewater annually [1]. The latex

wastewater contains high organic and high

nitrogen (COD: 1,000 ÷ 10,000 mg/l, BOD5:

1,700 ÷ 9,000 mg/l and total nitrogen: 45 ÷

1,600 mg/l) [2]. To meet Vietnam Industrial

effluent standards, the reuse of secondary

treated effluent for rubber tree irrigation may

be a potential benefit in terms of treatment

cost reduction as well as nutrient reuse. Due

to high nutrient composition, orientation to

reuse wastewater after aerobic biological

treatment for irrigation combining with higher

processing soil treatment may reduce costs of

chemical and electricity. In addition, reuse of

nutrient compositions (nitrogen, phosphorous,

potassium) may decrease the amount of


chemical fertilizers and improve soil quality

by providing useful microorganisms and

humus after aerobic biological treatment. The

study aimed to assess nutrient contamination

to aquifer from reuse of latex wastewater for

perennial tree irrigation.


Experimental site

The experiment was conducted at the Rubber

Research Institute of Vietnam (RRIV) in Ben

Cat District, Binh Duong province, Vietnam.

A pilot experiment (Figure 1) was done in

two lots of 100 rubber trees each. The

distance between rubber tree of each lot was

7m x 2.5m. Lot 0 (L0) was controlled without

irrigation. Lot 1 (L1) was watered by latex

wastewater after anaerobic stage with 20 liters

per rubber tree. Two observation wells about

18-20 meters in depth were drilled in the two

lots L0, L1.


112

X X X X X X . . . X X X X X X L0: without

irrigation 7m

X X X X X X . . . X X X X X X

X X X X X X . . . X X X X X X L1: watering

with treated

latex wastewater 7m

X X X X X X . . . X X X X X X

Rubber

trees Well 1 Well 0

Figure 1. Pilot scale irrigation experiment

The used irrigation water was taken from

effluent of a facultative waste stabilization

pond of latex processing industry. Hydraulic

rate, nutrient loading and COD loading

applied to the lot were 8m3/ha/week, 12.5

kgN/ha/day and 0.5 kg COD /ha/day

respectively.

Sampling and testing quality of the

wastewater

Treated wastewater was irrigated on every

Thursdays, from 08:00am during 1-2 hours.

Irrigation flow at each rubber’s root was

controlled by a counter clock to ensure 20

liters in every 12-14 seconds per rubber tree,

or equal to 10 m3 per ha per week. The

experiment was conducted in 16 weeks.

Water samples from the 2 observation wells

were taken by water pumps and then brought

to the laboratory for analyzing pH, COD,

TKN, N-NH4, N-NO3 and P - PO4.

Table 1. Characteristics of latex wastewater after

anaerobic pond

No Latex

wastewater

Average

concentration

1 pH 6.77

2 COD (mg/L) 407

3 TKN (mg/L) 176

4 NH3 – N (mg/L) 157

5 PO4 – P (mg/L) 282

Table 2. Modeling of latex wastewater

Type of

stream,

(i)

Latex wastewater (CS)

Latex

wastewater

of

anaerobic

pond

Anaerobic

effluent

Effluent of

facultative/

secondary

treatment

lt kk hk

COD (j),

mg/L 1000 500 250

TKN (k),

mg/L 500 150 60

Irrigation

load TL

(n)

Irrigation load m3/ha/day

30

50

80

120

Note: (i) – Type of wastewater: bi, kk, hk; (j) – COD: 1000, 500, 250 mg/L; (n) – 30, 50, 80, 120 m3/ha/day

Soil sampling and analyzing

Surface soil were sampled for SEM analysing at two sites: i) Inside the lot L0, soil were taken from the top, at 35cm depth and then mixed for one sample named D3; and ii) Inside the lot L1, soil were taken from the top, at 25cm depth and then mixed for one sample named CS3. SEM images at magnification 100 and 1,000 were conducted at the Innovation lab, Ryerson University, Toronto, Canada.

Input data for Groundwater Modeling System (GMS)

Irrigation load calculation option (n) 30, 50, 80 and 120 m3/ha/day, the results of the water analysis by the software MODFLOW with input data showned in the Table 2.


113

RESULTS AND DISCUSSIONS

Pilot scale irrigation experiment

Static water levels in observation wells were

at -9.5m as the drilling report while the depth

of exploited water were about -20 m. The

control experiment provided undetectable P -

PO4 presence. Phosphorus is usually in form

of phosphate in acidic soils (alkaline soils) in

which phosphate is adsorbed by iron and

aluminum oxide. The effect of soil

phosphorus adsorption is high if using

wastewater rubber with much phosphorus.

Concentrations of COD, TKN, N – NH4 of

the observation well No. 1 were lower than

the control well’s. These results showed

similar to the permeability experiment (Lam,

2013), at a depth of 20 m, TKN and N - NH4

were not detected, due to adsorption process

of soil together with biological conversion of

ammonia.

Pollution spreading in groundwater

The results of spread calculate by GMS

software are shown in table 3, which show

adversely distance from the border of

irrigation area (with sides 100m × 100m) to

the position at which the concentrations of

COD and ammonia reaching the limit values

of groundwater quality of QCVN

09:2008/BTNMT for edible use (COD ≤ 4

mg/l and N – NH3 ≤ 0.1 mg/l).

Note: Qmax = Irrigation load (m3/ha/day), H –

Static water level, m

When irrigating latex wastewater with low

concentrations of COD and ammonia (COD =

500 mg/l, N – ammonia = 150 mg/l), wells

must be 60 m far from irrigated areas. This

distance requires 30 m if COD is 250 mg/l.

Results of COD and BOD of the GMS

showed similar to the report of the EPA

(1981), water quality by land treatment with

BOD5 ≤ 5 mg / l and N-NH4 < 2 mg/L for the

irrigation load weight 6 cm / week and BOD5

load from 7-35 kg BOD5/ha/week.

a) b) c)

Figure 2. Change of contaminant concentrations: (a) COD, (b) TKN and (c) N – ammonia for irrigation

water and water observation wells

Table 3. The results of spread calculate by GMS software

Distance (d) from the borders adverse to

reach COD= 4 mg / l at static water level (m)

Distance (d) from the borders adverse to reach

ammonia= 0.1 mg / l at static water level (m)

Rubber

Waste

water

H(m)

Qmax

COD

(mg/l)

0.5

30

2,0

50

5,0

80

10

120

H(m)

Qmax

N – NH3

(mg/l)

0,5

30

2,0

50

5,0

80

10

120

1000 60 80 80 100 500 10 10 10 30

500 30 60 60 80 150 10 10 10 20

250 10 30 40 60 60 10 10 10 10


114

Scanning electron microscopy SEM

Scanning electron microscopy (SEM) images

was used to compare the soil quality among

controlled site (D3) and irrigated site (CS3),

detailed in Fig.3. At magnification 100, the

images showed that the porous medium of

CS3 had a lower total porosity than D3.

However, soil porosity between CS3 and D3

were not different at magnification 1000 and

may result in contaminants movement in the

same way among these sites.

CONCLUSION REMARKS

The results of latex wastewater spreading

showed that shallow groundwater was

affected with COD and ammonia in the

irrigation areas higher than QCVN

09:2008/BTNMT, column A. This

groundwater should be used for industry or

irrigation targets. Scale irrigation experiment

in the pilot at rubber block with the amount of

water for the experiment block is about 8

m3/ha/week and the load of COD water is 0,5

kg COD/ha/day with latex effluent, bCOD

absorbed in static water 9.5 m can be reduced

completely (~ 100%). As results of the GMS,

TN and COD concentrations spread in the

groundwater have stable after 10 months of

continuous irrigation with static water levels 2

m and after 12 months with static water levels

10 m.

FUTURE WORKS

Nutrients of the treated latex processing may

be reused for perennial trees like rubber trees.

However, impacts of the reclaimed water

need to be evaluate toxicity of chemical

substances and mixtures using natural soils to

microbial populations indigenous. The

specific objectives are to assess the changes

in soil chemical and physical properties

induced by irrigation, to highlight the

involved microbial health, and, to

characterize the role and behavior of the

organic matter.

Acknowledgment. The authors deeply

appreciate financial support of DOST of Binh

Duong.

a) b)

a)

c) d)

Figure 3. SEM at magnification of 100: a) D3 – inside L0, b) CS3 – inside L1,

and 1000: c) D3 – inside L0, d) CS3 – inside L1


115

REFERRENCES

1. Vietnam Rubber Group – Report of

wastewater treatment 2010 (2010),3 (in

Vietnamese)

2. Viet. N. T – Thesis: Sustainable treatment of

rubber latex processing wastewater: The UASB

system combined with aerobic post-treatment,

Proefschrift Wageningen Universiteit, 12.

3. Dan N. P, Nhung T. T. M, An H. K –

Establishing standard and management

mechanism of reclaimed water. Environment and

Natural Resources Protection Conference 2006-

2009. DOST - HCMC, 12/2009. (in Vietnamese)

4. Shiklomanov. Igor.A, World water resources –

A new appraisal and assessment for the 21st

century, United Nations, Educational, Scientific

and Cultural Organization, 1998, 27.

5. ANPHA, AWWA, WPCF - Standard methods

for the Examination of Water and Wastewater,

20th Edition. American Public Health Association.

Washington, D.C. (1999).

6. U.S. EPA. Wastewater Technology Fact Sheet

Rapid Infiltration Land Treatment. EPA 832-F-02-

12, September 2002, 3.

7. Cincinnati., Ohio, Process design manual land

treatment of Muicipal wastewater effluent,

EPA/625/R-06/016, September 2006,2.

8. QCVN 01:2008/BTNMT: National technical

regulation on the effluent of natural rubber

processing industry.

9. QCVN 09: 2008/ BTNMT: National technical

regulation on underground water quality.

10. Peng Wang and Arturo A. Keller. Natural and

Engineered Nano and Colloidal Transport: Role of

Zeta Potential in Prediction of Particle Deposition,

Langmuir 2009, 25(12), 6856–6862.

11. Yong Wang and Jun Jiang. Phosphate

adsorption at variable charge soil/ water interfaces

as influenced by ionic strength. Australian Journal

of Soil Research, 2009, 47, 529-536.

TÓM TẮT

ĐÁNH GIÁ TÁI SỬ DỤNG NƯỚC THẢI NGÀNH CHẾ BIẾN MỦ CAO SU

ĐẾN CHẤT LƯỢNG NƯỚC NGẦM CHO TRỒNG CÂY LÂU NĂM

Đặng Vũ Xuân Huyên1,*, Đặng Vũ Bích Hạnh1,

Amira Abdelrasoul2, Huu Doan2, Nguyễn Phước Dân1 1Đại học Bách khoa Thành phố Hồ Chí Minh – Việt Nam

2Đại học Ryerson, Toronto, Canada

Nghiên cứu nhằm đánh giá ô nhiễm nước ngầm trong tái sử dụng từ nước thải mủ cao su để tưới

cây lâu năm. Nước thải mủ cao su có chứa hàm lượng nitơ cao và BOD cao, đòi hỏi phải được áp

dụng công nghệ xử lý bậc cao để có thể đáp ứng tiêu chuẩn nước thải công nghiệp Việt Nam. Việc

tái sử dụng nước thải sau xử lý bậc hai để tưới cây cao su có thể mang lại lợi ích do giảm chi phí

xử lý đồng thời tái sử dụng chất dinh dưỡng.

Mô hình thí nghiệm được tiến hành với hai nghiệm thức, đó là nghiệm thức tưới và nghiệm thức

không tưới với 100 cây cao su cho mỗi nghiệm thức. Diện tích mỗi nghiệm thức là 7m x 2,5m.

Nước tưới được lấy từ đầu ra sau bể ổn định của hệ thống xử lý nước thải mủ cao su. Lưu lượng,

tải lượng ô nhiễm và COD được sử dụng là 8m3/ha/tuần, 12,5 kgN/ha/ngày và 0,5kg

COD/ha/ngày.

Mô hình GMS đã được áp dụng để đánh giá lan truyền nitơ và COD trong tầng nước ngầm. Mô

hình cho thấy phản ứng chuyển hóa sinh học của COD và nitơ trong tầng vadose. Kết quả nghiên

cứu cho thấy tổng nitơ và COD trong nước ngầm đạt được mức độ ổn định ở độ sâu 2 mét sau khi

tưới 10 tháng và ở độ sâu 10 mét sau khi tưới 12 tháng. Ngoài ra, kính hiển vi điện tử quét (SEM)

được sử dụng để so sánh chất lượng đất giữa các vị trí thử nghiệm.

Từ khóa: Nước thải nganh chế biến mủ cao su, tái sử dụng dinh dương, tiêu chuẩn nước thải công

nghiệp Việt Nam, mô hình dong chảy nước dưới đất (GMS), kinh hiển vi điện tử quet (SEM).



116


117

ASSESSMENT IMPACT OF RECLAIMED WATER TO SOIL QUALITY

BASED ON SOIL MICROBIAL COMMUNITY TOXICITY

Huyen Vu Xuan Dang*, Huyen T Bich Trinh,

Hanh Vu Bich Dang, Dan Phuoc Nguyen Hochiminh city University of Technology, VNU-HCM, Vietnam

ABSTRACT

The research aimed to evaluate environmental risk of reused wastewater on soil applying slow-rate

land treatment to irrigate rubber field. The experiment was conducted in the Vietnam Rubber

Research Institute, Ben Cat ward, Binh Duong province. The wastewater of latex processing after

secondary treatment may potentially harm to soil quality that would be showed through microbial

community. The method used natural bacteria and determined inhibition effect on the soil

microbial community bacteria in order to assess the changes in soil chemical and physical

properties induced by irrigation. NH3, and NO3 concentrations to determine microbial health

through measuring ammonification, nitrification, respectively, as a measure of the soil microbial

community to decompose organic matter and release plant nutrients every 7 days of exposure

during 28 days. Firstly, toxicity of the wastewater with different dilution (100%, 50%, 25% and

12.5%) causes soil microbial community activities and development as their nutrients. After

exposing to the reused wastewater at different concentrations, most of soil parameters were

increased at the day 7th. Until the day 28th of the experiment, concentrations of tested parameters

were still not decreased. Concentration of NH3 was decreased while organic carbon and total

aerobic microorganism increased that may be resulted from metabolism processes of the soil but

not from wastewater exposure. The reused wastewater of latex processing may be used to irrigate

rubber trees without inhibition to the soil health.

Keywords: Inhibitory effect, slow-rate land treatment, latex wastewater, toxicity test.

INTRODUCTION*

Vietnam to become the fourth largest exporter

of natural rubber in the world (800,000 ton

productivity) 4. At present, the growing

rubber area is spread from the northern to the

central provinces (less than 10,000ha) while

other west-southern areas from over 50,000ha

or between 10,000-50,000ha 5. The Vietnam

rubber group reported that the rubber industry

discharges 10 million m3 wastewater every

year. An average of loading rate of rubber

industry is 25 m3 wastewater/ton dried rubber,

35 m3/ton rubber product and 18 m3/ton latex.

Rubber wastewater contains high

contaminants, COD may be up to 1,000 –

10,000 mg/l, BOD5 may be 1,700 – 9,000

mg/l and total nitrogen may be 45 –


1,600mg/l (Viet, 1999). Wastewater of latex

processing was researched for irrigation 78

but pollutants are spread over a large area and

may affect the crops 6. A serious threat of

rubber wastewater towards environmental

protection is high concentration of nitrogen in

this effluent (Table 1).

Table 1. Characteristics of process effluents from

rubber processing 9

Parameter Typical range

pH 3.7 - 5.5

Biological oxygen demand 1,500 - 7000

Chemical oxygen demand 3500 - 14000

Suspended solids 200 - 700

Total nitrogen 200 - 1800

Sulphate 500 - 2000

All units are mg/l, except pH.


118

Identification of bacteria which can grow in

the concentrated latex wastewater was studied

by Choorit et al., 2003. After 40 h of

cultivation, 34% of COD was decreased by

Rubrivivax gelatinosus and Thiobacillus sp.

(Choorit et al., 2003). Four kinds of

Thiobacillus sp. were isolated from domestic

and rubber wastewaters in Thailand by

Kantachote and Innuwat (2004). All isolates

could grow in pH of 2.0 - 7.0 (optimum 6.5),

temperature of 25 - 45°C (optimum 30 -

35°C) under both aerobic and anaerobic

conditions. The highest COD removal (54%)

can be obtained by Thiobacillus sp. WI1

cultivated in rubber wastewater for 14 days

while the efficiency of strain WI4 for BOD

and COD removal was 83% and 46%

(Kantachote and Innuwat, 2004).

Microbial community in soil may improve the

quality of the soil. Assessment of nutrient

could show toxicity reduction of soil through

microbial community. The specific objectives

were to assess the changes in soil chemical

and physical properties induced by irrigation,

to highlight the involved microbial health,

and, to characterize the role and behavior of

the organic matter. NH3, and NO3-

concentrations (concentrations per gram of

soil) determined microbial health through

measuring ammonification, nitrification,

respectively, as a measure of the soil

microbial community to decompose organic

matter and release plant nutrients.


Experimental site

The experiment sites were rubber plantation

field located at the Vietnam Rubber Research

Institute (RRIV), Ben Cat ward, Binh Duong

province. The sites include rubber field and

rubber factory inside the RRIV. These areas

were under tropical climate. The annual

rainfall of this area is 1,800mm/year.

Soil sampling and toxicity test

The surface soils were sampled randomly at

0-15 cm depth to test their physical and

chemical properties using a core sampler. Soil

collection should not receive any fertilizer or

pesticide applications within the past 24

months. The moisture of the soil was 60%.

After collecting, the surface soil samples were

allowed to air dry until sievable by 2mm

mesh sieve.

Soil microbial community toxicity was

determined by Soil Microbial Community

Toxicity Test Guidelines US EPA 712-C-96-

161, 1996. The soil samples were exposed

with treated rubber wastewater at different

concentrations (100%, 50%, 25%, and

12.5%). The quality was tested every 7 days

during 28 days.

LABORATORY ANALYSIS

Soil chemical properties

Soil chemical properties were analyzed in the

Laboratory of Faculty of Environment and

Natural Resources of HCMUT including: soil

pH, nitrogen (NH3, NO2 , NO3, TKN), organic

matter, total aerobic microorganism.

Parameters were tested by Standard Methods

for the Examination of Water and Wastewater

(APHA, 1998), published by American Public

Health Association, American Water Works

Association and Water Environment

Federation.

Soil Microbial Community Toxicity Test

Soil microbial community toxicity was

determined by Soil Microbial Community

Toxicity Test Guidelines US EPA 712-C-96-

161, 1996. After sampling, all soil samples

were incubated in darkness at approximately

22oC. Soils are then sampled on 5th day and

28th day and analyzed for NH3 and NO3

concentrations to determine microbial health

through measuring ammonification,

nitrification, respectively. Control samples

were received a similar amount of water

without the reused wastewater.


119


Effects of the exposure to physical

properties of the soil

pH:

Then soil samples were exposed from the day

zero to the 28th. The result showed that pH of

all samples decreased from time to time, from

7 to 4 (Fig. 1). However, pH of control

samples were also down over time. This

proved that soil pH was not only effected by

contaminants of the wastewater but also

depended on substances of natural soil. This

report agrees with the observations of Eneje

(2012) who reported that the soil pH of the

rubber plantation is very strongly acidic

(4.38) at 0-15cm 12.

Moisture:

At wastewater concentrations of 12.5; 25; 50

and 100%; moisture decreased over time from

65.33% to 40.88% (Fig. 2). Reduction of

moisture from time to time affected on soil

pH. This agrees with the observation of Eneje

(2012) who reported that there was position

correlation between moisture and pH value.

Effects of the exposure to chemical

properties of the soil

Nitrogen:

Fig. 3 showed that TKN ratio of initial and

exposure samples were down from time to

time. With control samples, TKN

concentration was also down at lower level

than exposed samples. Being exposed at

different concentrations, variation of TKN

levels were very clearly at negative

correlation between wastewater

concentrations and soil TKN. This was

proved that low concentration of wastewater

supplied organic matter for microbes’

activities.

Figure 1. Variation of pH during the exposure of

reused wastewater with different dilution

Figure 2. Variation of moisture during the exposure

of reused wastewater with different dilution

Figure 3. Variation of TKN during the exposure of


Figure 4. Variation of NH3 during the exposure of



120

NH3 was clearly correlated with soil

microbial community activities. Fig. 4

showed that NH3 of control samples was

slightly lower than other exposed samples.

The result showed that treated rubber

wastewater remained nutrients supplied to

microbes development. However, microbes’

activities were highest at 12.5% among other

concentrations. With 100% of wastewater

exposure, NH3 concentration was up at the

day 7th, and then decreased rapidly after the

day 14th. This was showed that contaminants

of the wastewater at high level inhibited soil

microbial community along with time.

Toxicity of the wastewater with different

concentrations causes soil microbial

community activities and development as

nutrients that were showed by NH3

increasing. In the next 14 days, combination

between nutrient reduction and soil

accumulative chemicals (metabolism products

of microbes) may inhibit on microbes that

was proved by NH3 decreasing. The decrease

was showed clearly in 100% compared with

NO2 incease to the 14th day at the same

concentration while NO3- increased until the

day 14th, 28th; NO2- decreased in negative

correlation with NO3- (Fig 5 and 6). At first,

microbes’ activities were in adaption stage

then proceed of the development and

parameters were up respectively. The higher

level of the wastewater may resulted in the

stronger inhibitory of the microbial

community.

Figure 5. Variation of NO2 during the exposure of


Figure 6. Variation of NO3 during the exposure of


Carbon:

Figure 7. Variation of organic carbon during the

exposure of reused wastewater with different

dilution

Figure 8. Variation of total aerobic

microorganism during the exposure of reused

wastewater with different dilution


121

While organic carbon and total aerobic

microorganism were increased (Fig 7 and 8).

This may be due to microbes’ development

creating biomass growth up. Thus,

concentration of NH3 reducing may not result

from higher level of the wastewater but from

other metabolism processes. In addition, there

was still no sign of reducing of its growth

until the day 28th. In conclusion, the reused

wastewater of latex processing may be used

to irrigate rubber tree and have no inhibition

to the soil health.

CONCLUSION

Toxicity test basing on soil microbial

community was quickly, easily to conduct for

toxic substances whose exposure is not

anticipated. This test can be used to measure

of the soil microbial community to

decompose organic matter and release plant

nutrients. After exposed to the reused

wastewater at different concentrations, most

of soil parameters were increased at the day

7th. Until the day 28th of the experiment,

concentrations of tested parameters were still

not decreased. Concentration of NH3 was

decreased while organic carbon and total

aerobic microorganism increased that may be

resulted from metabolism processes of the

soil but not wastewater exposure. The reused

wastewater of latex processing may be used

to irrigate rubber trees without inhibition to

the soil health.

Acknowledgment. The authors deeply

appreciate financial support of DOST of Binh

Duong.

REFERENCES

1. Sun SF (2004). Physical Chemistry of

Macromolecules: Basic Principles and Issues.

John Wiley & Sons, INC Publication, pp. 150-

164.

2. Lonholdt J, Andersen LS (2005). Water and

Wastewater Management in the Tropics. IWA

Publishing, pp. 333-345

3. Xiaofei Z (2008). A Study on countermeasure

for labor force shortage in natural rubber industry

with reference to China Asean region integration.

In: Intl Conf China Asean Regional Integration:

Political Economy of Trade, Growth and

Investment, Uni. of Malaya.

4. Development of rubber trees in Vietnam,

Vietnam Sci. and Tech. Info. J., 2013 (3), 4-10.

5. Z. Li, J.M. Fox, Mapping rubber tree growth in

mainland Southeast Asia using time-series

MODIS 250 m NDVI and statistical data,

Applied Geography 2012, 32 (2), 420-432

6. B. Van der Bruggen, Chap. 3 The Global

Water Recycling Situation, Sus. Sci. Eng. 2010, 2,

41-62

7. B.N. Nguyen, Recycled wastewater of rubber

effluent for watering Guinea grass Panicum

maximum, The final report for Rubber Research

Institute of Vietnam, 2000.

8. B.N. Nguyen, Rubber processing wastewater

for irrigation of rubber plantations, J. Agri Rural

Develop. 2007, (15), 88-90.

9. Mitra Mohammadi, Hasfalina Man, Mohd Ali

Hassan and Phang Lai Yee. Treatment of

wastewater from rubber industry in Malaysia;

African J. of Biotech. Vol. 9(38), pp. 6233-6243,

2010

10. Statistical YearBook of Vietnam. Statistical

Publishing House of Vietnam, 2013

11. Rungruang N, Babel S. Treatment of natural

rubber processing wastewater by combination of

ozonation and activated sludge process. In:

International Conference on Environmental

Research and Technology (ICERT 2008),

Malaysia, pp. 259- 263.

12. Eneje, R.C. nad Apundu J. Soil water and

inonic properties, and C:N ratio of plantation soils

in Micheal Okpara university of agriculture

umudike Abia state. Intl. J. of Agri. Sci. and

Biores Eng. Res. Vol. 1(2), pp. 51-58, 2012.


122

TÓM TẮT

ĐÁNH GIÁ ẢNH HƯỞNG NƯỚC TÁI SINH ĐẾN CHẤT LƯỢNG ĐẤT

DỰA TRÊN ĐỘC TÍNH CỘNG ĐỒNG VI SINH VẬT ĐẤT

Đặng Vũ Xuân Huyên*, Trịnh Thị Bích Huyền, Đặng Vũ Bích Hạnh, Nguyễn Phước DânĐại học Bách khoa Thành phố Hồ Chí Minh

Nghiên cứu này nhằm đánh giá rủi ro môi trường của nước thải được tái sử dụng bằng phương

pháp cánh đồng lọc chậm để tưới cho cây cao su. Thí nghiệm được tiến hành tại Viện Nghiên cứu

Cao su Việt Nam, phường Bến Cát, tỉnh Bình Dương. Nước thải của ngành chế biến mủ cao sau

xử lý thứ cấp có khả năng gây tổn hại đến chất lượng đất được đánh giá qua hoạt động của cộng

đồng vi sinh vật. Phương pháp này sử dụng vi khuẩn tự nhiên và xác định khả năng ức chế vi

khuẩn trên cộng đồng vi sinh vật đất để đánh giá những thay đổi về hóa học và tính chất vật lý đất

do nước tưới gây nên. Nồng độ NH3, NO3 được sử dụng để xác định sức khỏe của vi sinh vật

thông qua các thông số quá trình amon hóa, nitrat hóa. Phương pháp này đánh giá hoạt động phân

hủy chất hữu cơ và giải phóng các chất dinh dưỡng của cộng đồng vi sinh vật đất mỗi 7 ngày tiếp

xúc trong thời gian 28 ngày. Đầu tiên, cộng đồng vi sinh vật sử dụng các chất có trong nước thải ở

các nồng độ pha loãng khác nhau (100%, 50%, 25% và 12,5%) để hoạt động và phát triển. Sau

một thời gian tiếp xúc với nước thải, hầu hết các thông số đất đều tăng lên ở ngày thứ 7. Cho đến

ngày thử nghiệm 28, nồng độ của các thông số thử nghiệm vẫn không giảm. Nồng độ NH3 giảm

trong khi cacbon hữu cơ và tổng vi sinh hiếu khí tăng có thể là kết quả của quá trình trao đổi chất

của đất, không phải do tiếp xúc với nước thải. Nước thải tái sử dụng chế biến mủ cao su có thể

được sử dụng để tưới cho cây cao su mà không có sự ức chế đối với sức khỏe đất.

Từ khóa: Ức chế, phương pháp cánh đồng lọc chậm, nước thải nganh chế biến mủ cao su, thử

nghiệm độc tinh cộng đồng vi sinh vật đất.


Tạp chí Khoa học và Công nghệ

SỐ ĐẶC BIỆT CHÀO MỪNG 49 NĂM THÀNH LẬP TRƯỜNG ĐẠI HỌC KỸ THUẬT CÔNG NGHIỆP

(19/8/1965 – 19/8/20140

Mục lục Trang

Phan Quang Thế, Nguyễn Thị Quốc Dung, Phan Thị Phương Thảo - Ảnh hưởng của độ cứng phôi trong quá

trình tiện cứng thép hợp kim 3

Nguyễn Văn Lanh, Đàm Bảo Lộc - Hệ thống thích nghi mô hình mẫu trực tiếp dựatrên hệ thống điều khiển thích nghi cho hệ thống twin rotor momo 9

Nguyễn Minh Ý, Phạm Ngọc Thăng, Nguyễn Huy Toán - Một phương pháp tiếp cận mới cho việc tiết kiệm điện

năng cho các hộ tiêu thụ dựa trên công nghệ Smartgrid 15

Gia Thị Định, Trần Mạnh Tuấn, Trần Quế Sơn - MRAS trực tiếp với các điều kiện an toàn ứng dụng điều khiển robot di động hai bánh 21

Lê Tiên Phong, Ngô Đức Minh - Nghiên cứu thiết kế hệ thống quản lý năng lượng cho nguồn pin mặt trời độc lập 29

Lê Tiên Phong, Nguyễn Thị Mai Hương, Nguyễn Tiến Hưng - Điều khiển hệ thống pin mặt trời kết nối lưới sử

dụng bộ biến đổi flyback đan xen 37

Nguyễn Hoài Nam, Trịnh Thị Minh Nguyệt - Một tiến trình đào tạo mới cho một phân lớp của các mạng rron

hồi quy 43

Nguyễn Thị Hồng Cẩm, Nguyễn Văn Trang, Vũ Ngọc Pi - Nghiên cứu mới về tối ưu hóa tỷ số truyền các bộ

truyền trong hộp hành tinh 47

Nguyễn Thị Thu Thủy - Xây dựng dữ liệu bộ hệ số phát thải các chất ô nhiễm không khí trong quá trình cán thép 51

Nguyễn Tiến Duy, Lâm Hoàng Bình, Lâm Hùng Sơn, Nguyễn Phương Huy - Ứng dụng đại số gia tử để thiết

kế bộ điều khiển lượng oxy hòa tan trong quá trình xử lý nước thải theo phương pháp bùn hoạt tính 57

Phạm Thị Minh Khuyên, Phạm Thị Mai Yến - Quản trị chuỗi cung ứng giáo dục: giải pháp nâng cao hiệu quả

các chuyển giao khoa học công nghệ 63

Nguyễn Đức Thịnh, Phan Thị Phương Thảo, Nguyễn Thị Lâm Oanh - Thiết kế và chế tạo quét nhà thông minh 69

Nguyễn Đức Thịnh, Vũ Anh Tuấn, Nguyễn Văn Dự - Thiết kế và chế tạo động cơ nhiệt stirling sử dụng năng

lượng mặt trời 73

Phan Thị Thu Hà, Phan Thị Phương Thảo - Ảnh hưởng của nhiệt độ đến đặc tính biến dạng tốc độ cao của vật

liệu composite graphene nền polyurethane 77

Trương Thị Hồng Thúy, Đỗ Thị Hồng Nga - Ứng dụng của mạng Neural trong chẩn đoán bệnh viêm gan 81

Vũ Ngọc Huy, Trần Mạnh Tuấn, Nguyen Thị Mai Hương, Nguyễn Tiến Hưng - Điều khiển số động cơ một chiều qua mạch thyristor công suất 87

Vũ Ngọc Huy, Trần Mạnh Tuấn, Nguyen Thị Mai Hương, Nguyễn Tiến Hưng - Điều khiển bền vững động cơ

một chiều 93

Vũ Ngọc Kiên, Đào Huy Du, Nguyễn Hữu Công - Tthuật toán giảm bậc bảo toàn điểm cực dựa trên phân tích Schur 101

Vũ Quốc Việt - Nâng cao hiệu quả quá trình loại bỏ thạch tín trong nước sinh hoạt 107

Đặng Vũ Xuân Huyên, Đặng Vũ Bích Hạnh, Amira Abdelrasoul, Huu Doan, Nguyễn Phước Dân - Đánh giá

tái sử dụng nước thải ngành chế biến mủ cao su đến chất lượng nước ngầm cho trồng cây lâu năm 111

Đặng Vũ Xuân Huyên, Trịnh Thị Bích Huyền, Đặng Vũ Bích Hạnh, Nguyễn Phước Dân - Đánh giá ảnh

hưởng nước tái sinh đến chất lượng đất dựa trên độc tính cộng đồng vi sinh vật đất 117

Journal of Science and Technology

127(13)

N¨m 2014

Documents

So 127 (13) 12-2014.cdr