Upload
vuthu
View
225
Download
0
Embed Size (px)
Citation preview
Hindawi Publishing CorporationJournal of OptimizationVolume 2013 Article ID 142390 11 pageshttpdxdoiorg1011552013142390
Research ArticleOptimization of Integer Order Integrators for DerivingImproved Models of Their Fractional Counterparts
Maneesha Gupta and Richa Yadav
Advanced Electronics Lab Department of ECE Netaji Subhas Institute of Technology Dwarka New Delhi 110078 India
Correspondence should be addressed to Richa Yadav yadavricha1gmailcom
Received 28 February 2013 Revised 21 May 2013 Accepted 30 May 2013
Academic Editor Manuel Lozano
Copyright copy 2013 M Gupta and R Yadav This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited
Second and third order digital integrators (DIs) have been optimized first using Particle SwarmOptimization (PSO)withminimizederror fitness function obtained by registering mean median and standard deviation values in different random iterations Laterindirect discretization using Continued Fraction Expansion (CFE) has been used to ascertain a better fitting of proposed integerorder optimized DIs into their corresponding fractional counterparts by utilizing their refined properties now restored in themdue to PSO algorithm Simulation results for the comparisons of the frequency responses of proposed 2nd and 3rd order optimizedDIs and proposed discretized mathematical models of half integrators based on them with their respective existing operators havebeen presented Proposed integer order PSO optimized integrators as well as fractional order integrators (FOIs) have been observedto outperform the existing recently published operators in their respective domains reasonably well in complete range of Nyquistfrequency
1 Introduction
In fractional calculus (FC) initially the development ofmathematical models of fractional order integrators (FOIs)and fractional order differentiators (FODs) [1 2] startedpresumably for searching different generalizing approachesfor switching from integer order to the fractional orderdomain The frequency response of ideal fractional orderdiffer-integrator (119904
plusmn120572) is
119867fractional (119895120596) = (119895120596)plusmn120572
(1)
where 119895 = radicminus1 and 120596 gives angular frequency in radiansVariable ldquo120572rdquo defines the order of fractional order operatorsand its value lies in between 0 and 1
Due to its exotic nature FC has strengthened its graspover different research areas such as statistical modelingmechanical system analysis [3] control [4] automated con-trol [5] instrumentation [6] signal processing [7] and radioengineering and image processing [8 9] At present timeFC is also extensively spreading its niche in different design
methods for improving fractional order controllers by frac-tional integral and derivative functions [10 11] An importantdesign method which has undoubtedly helped in sharpeningup the focus of an efficient formulation of new blocks offractional operators is ldquolinear interpolationrdquo method [12ndash16] This most vastly followed tri-step method involves thefollowing steps (1) First step deals with linear interpolationof two existing digital integrators (2) In second step a newfractional order integrator is formulated by discretization ofinteger order integrator developed in first step by anyone ofthe two existing discretization (direct and indirect) schemes(3) In the third step this expanded series is truncatedup to some arbitrary order (119873) which defines the orderof the new fractional operator (119904
plusmn120572) Direct discretization[17] works with the help of different expansion techniquessuch as Taylor Series Expansion (TSE) [18] Power SeriesExpansion (PSE) [19 20] and Continued Fraction Expansion(CFE) [21 22] whereas for indirect discretization [23ndash26]different rational approximations have been developed bymany mathematicians namely Carlson and Halijak [27]Steiglitz [28] Khovanskii [29] Roy [30] Oustaloup et al [31]
2 Journal of Optimization
andMaione [32] which proved like boon for the developmentof fractional integratorsdifferentiators
In spite of the efficient and systematic nature of dif-ferent previously mentioned approximations still thereexisted a dissent about performance of these approximationswhen these were later adapted by well-known discretiza-tion schemes for obtaining FOIs Moreover development ofnew approximations requiredmany exhaustivemathematicalapplications to make them perform in finite dimensionalform in proper range of frequencies during discretizationInstead of deriving new methods of rational approximationsauthors have suggested here an alternate solution in whichthe integrator operator (used as 119904-119911 transform) itself shouldbe pushed near boundaries of an optimal solution (near idealresponse) before converting it to fractional system Reasonbehind this is that every minute dispensation in propertiesof this integer order operator will directly propel into itsnoninteger order counterpart The main novelty proposed bythis paper is to put together unique combination of opti-mized integer order operators and an accurate approximationtechnique for configuring their efficient fractional order (fo)operators The current scenario of design methods for deriv-ing optimal operators [33ndash41] is cumulatively proceedingtowards ldquonot so distantrdquo phase of perfect optimization withnegligible errors So another motivating point behind theapproach proposed in this paper is to trespass all disadvan-tages of existing conventional design methods of FOIs byusing refined capabilities of an efficient optimization scheme
The frequency response of ideal integer order digitalintegrators (DIs) is
119867int (120596) = (1
119895120596) where |120596| le 120587 (2)
and the frequency response of ideal integer order digital dif-ferentiators (DDs) is
119867diff (120596) = 119895120596 where |120596| le 120587 (3)
where 119895 = radicminus1 120596 is the angular frequency in radians andldquo119879rdquo is the sampling period
In recent novel advancements different optimizationalgorithms namely Linear Programming (LP) [33 34]Genetic Algorithm (GA) [35ndash37] Simulated Annealing (SA)[38 39] and pole-zero (PZ) optimization [40 41] havebeen used for obtaining more refined integer order differ-integrators Jain et al [35] have recently derived optimalmodels of recursive DIs and DDs by GA algorithm whileUpadhyay [41] has developed these operators by analyzingpoles and zeros of existing recursive DDs by PZ optimizationtechnique for obtaining better results for frequency responsesover wideband of complete frequency spectrum Al-Alaoui[39] has done a credible work in improving design processof DIs and DDs based on Newton-Cotes integration rules byusing linear interpolation and SA
This paper attempts to find the optimal solutions ofDIs by using an efficient modified application of Parti-cle Swarm Optimization (PSO) technique [42ndash44] beforediscretizing them to their fractional mathematical models
Values of mean median and standard deviation in differentrandom iterations have been continuously registered usingPSO for the optimization of 2nd and 3rd order DIs withminimized error fitness function Later these optimized DIsare discretized by CFE of indirect discretization scheme [23]for finding half integrator models of different orders Thesimulation results of magnitude responses phase responsesand relative magnitude errors for the proposed optimizedDIs have been compared with all the recently publishedintegrators optimized by different optimization algorithmsnamely LP [34] GA [35] SA [39] and PZ [41] Frequencyresponses of the proposed FOIs based on 2nd and 3rd orderoptimized DIs have been also compared with their existingfractional models [2 17] in complete frequency range
The paper is organized as follows Section 2 deals withthe brief description of original and modified PSO algo-rithm for optimizing DIs along with its trade-off con-ditions and the resultant 2nd and 3rd order optimizedintegrators thus derived have been presented in this sec-tion Comparisons of proposed and existing DIs are alsogiven in the same section These optimized operators havebeen discretized by indirect discretization using CFE tech-nique of indirect discretization for deriving models of FOIsin Section 3 Section 4 presents the simulation results ofcomparisons of proposed half integrators with responsesof ideal and the existing models Section 5 concludes thepaper
2 PSO Algorithm for Optimizing 2nd and 3rdOrder Integer Order Integrators
21 Outline of Basic Functionality of Original PSO AlgorithmIn conventional PSO algorithm [42ndash44] first a popula-tion of random solutions has been initialized and differentgenerations are updated for searching optimal values Inthis algorithm the potential solutions called particles flythrough the problem space by following the current optimumparticles The position of each particle encodes a possiblesolution to the given problem The velocity of the particle isthe parameter to be added to the current position to find thenewposition of the next generationThe velocity and positionare updated for all particles in every generation unless anoptimal solution is found The particles keep track of itscoordinates in the problem space which are associated withthe best solution (fitness) it has achieved so far These valuesare called pbest and the best element among all the pbestvalues of all the particles is called gbest The velocities areinitialized to zero and after finding the two best values theparticle updates its velocity and positions in accordance withthe following update equations
V (119894) = V (119894 minus 1) + 11988811199031
(119901pbest minus 119901 (119894)) + 11988821199032
(119901gbest minus 119901 (119894))
119901 (119894 + 1) = 119901 (119894) + V (119894)
(4)
where V(119894) is the current particle velocity V(119894 minus 1) is previousparticle velocity 119901(119894) is current particle position 119901(119894 + 1) isthe position of particle in next generation 119901pbest is the best
Journal of Optimization 3
Define the solution space fitness function and population size
Initialise position velocity
For each particle
Evaluate fitness function
Next particle
orNext iteration
Solution is gbest
Update velocityand position
Yes
No
If fitness (x) better than
pbest and gbest
fitness (gbest)gbest = x
No iteration = max no iteration
fitness (gbest) is good enough
minus |H(ej120596)|)2d120596
If fitness (x) better thanfitness (pbest)
pbest = xEint = int
120587
0(1120596
Figure 1 Flowchart of modified PSO algorithm
position the particle has seen 119901gbests is the best position theswarm has seen 119888
1 1198882are constants set to 2 and 119903
1 1199032are the
random numbers between 0 and 1
22 Parameters Used for Optimization and Trade-Offs betweenMultiple Parameters during Convergence of Optimized DigitalIntegrators Amathematical model of transfer function (TF)of optimized digital integrator (DI) can be described ingeneralized form as
119867 (119911) = (119861119872
(119911)119887119872 + sdot sdot sdot + 119861
1(119911)1198871 + 119861
0(119911)1198870
119860119873
(119911)119886119873 + sdot sdot sdot + 119860
1(119911)1198861 + 119860
0(119911)1198860
) (5)
where119860119894and 119861
119895 are arbitrary constants whereas 119886
119894and 119887119895are
real numbers with (119894 = 0 1 119873) and (119895 = 0 1 119872)
Transfer functions (TFs) of 2nd and 3rd order DIs whichhave been optimized using PSO in this paper are given in (6)and (7) respectively
119867PSO 2nd opt (119911) = (11986121199112 + 119861
1119911 + 1198610
11986021199112 + 119860
1119911 + 119860
0
) (6)
119867PSO 3rd opt (119911) = (1198613(119911)3
+ 1198612(119911)2
+ 1198611119911 + 1198610
1198603(119911)3
+ 1198602(119911)2
+ 1198601119911 + 119860
0
) (7)
The coefficients of the previously mentioned TFs are gener-ated by random iterations of PSO algorithm as it takes realnumbers as particles
4 Journal of Optimization
Table 1 Transfer functions and fitness values of worst and average cases of 2nd order PSO optimized DIs
TFs of worst cases of 2nd order DIs along with their fitnessfunction values (FV)
TFs of average cases of 2nd order DIs along with their fitnessfunction values (FV)
Transfer function FV Transfer function FV
1198671(119911) = (
03733241199112 + 0427523119911 + 0397575
1199112 minus 0042561119911 minus 0999916) 0643052 119867
1(119911) = (
0228961199112 minus 0825829119911 minus 0502494
1199112 minus 0776284119911 minus 0181729) 0382107
1198672(119911) = (
minus09292511199112 + 0381076119911 minus 075834
1199112 minus 0725805119911 minus 0306558) 390016 119867
2(119911) = (
minus03364781199112 minus 082224119911 + 0101197
1199112 minus 0818946119911 minus 0147828) 0407093
1198673(119911) = (
05416991199112 + 0241831119911 + 0506633
1199112 minus 00552516119911 minus 1) 473234 119867
3(119911) = (
008498351199112 + 0752164119911 + 0756218
1199112 minus 0167219119911 minus 0882869) 050102
1198674(119911) = (
minus03545831199112 minus 0833923119911 minus 0813869
1199112 minus 00637125119911 minus 1) 589578 119867
4(119911) = (
minus0397291199112 minus 0795585119911 minus 0130415
1199112 minus 0722962119911 minus 0318656) 0618717
1198675(119911) = (
03538521199112 + 0203374119911 + 0466221
1199112 minus 00406162119911 minus 1) 18799 119867
5(119911) = (
02843821199112 + 0949776119911 + 0363676
1199112 minus 0401952119911 minus 0648204) 0737534
Magnitude response of the optimized DI is
119867mag = |119867 (119908)| =abs (numerator)abs (denominator)
(8)
Phase response of the optimized DI is
119867phase = arg |119867 (119908)| (9)
Relative magnitude error (dB) or RME value is given by
RME = 20 log10
100381610038161003816100381610038161003816100381610038161003816
119867ideal (120596) minus 119867approx (120596)
119867ideal (120596)
100381610038161003816100381610038161003816100381610038161003816
= 20 log10
1003816100381610038161003816100381610038161003816100381610038161003816
(1119895120596) minus 119867approx (120596)
(1119895120596)
1003816100381610038161003816100381610038161003816100381610038161003816
(10)
where119867ideal(120596) is themagnitude of ideal value of the operatorand 119867approx(120596) is the value of the digital integrator that isapproximated
In the recently published literature on digital and frac-tional operators most of the researchers have focused onformulating the fitness function for optimization techniquebased on only single parameter [38ndash41] that is magnitudeerror because practically if phase and magnitude are opti-mized for a multiparameter problem the solution is alwaysstuck in a trade-off due to their contradictory nature Thereason behind this is that when the optimal solution startsconverging towards the best results of one parameter (mag-nitude response) it simultaneously degrades optimality ofsecond parameter (phase responses) after certain number ofiterations Solution for such trade-offs is to choose domi-nating parameter in problem under consideration In thispaper the design strategy to find optimized DIs is kept moreconcerned for minimum magnitude response error whichis comparatively more significant characteristic The fitnessfunction used to approximate a digital integrator is taken asthe mean square error of the transfer function and it is givenas
119864int = int120587
0
(1
120596minus
10038161003816100381610038161003816119867 (119890119895120596
)10038161003816100381610038161003816)2
119889120596 (11)
1 15 2 25Frequency
Mag
nitu
de re
spon
se
Comparison of magnitude responses of proposed DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
1
09
08
07
06
05
04
03
02
with existing optimized DIs
Figure 2 Comparison ofmagnitude responses of proposed 2nd and3rd order optimized DIs with existing [26 34 39 41] optimizedmodels
Aim of PSO optimization is to search the best optimalsolution (TF of DI) with minimized mean square error
23 Modified PSO Algorithm for Derivation of OptimizedDIs In this paper a slight modification in the originalPSO algorithm has been introduced in velocity taking intoaccount the given problem of deriving the best feasibleoptimized DIs with very small magnitude response errorswithout worsening their corresponding phase responses Inmodified PSO used in this paper the maximum velocitywhich constraints the movement of swarms to the range is
Journal of Optimization 5
0 05 1 15 2 25 3
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Comparison of RME values of proposed DIs
minus20
minus40
minus60
minus80
minus100
minus120
Al-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)Upadhyay (pole-zero optim)
Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
with existing optimized DIs
Figure 3 Comparison of RMEvalues of proposed 2nd and 3rd orderoptimized DIs with existing [26 34 39 41] optimized models
0
50
100
150
200
Frequency
Phas
e (de
g)
Comparison of phase responses of of proposed DIs
minus50
minus100
minus150
minus2000 05 1 15 2 25 3
with existing optimized DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
Figure 4 Comparison of phase responses of proposed 2nd and 3rdorder optimized DIs with existing [26 34 39 41] optimized models
divided by a factor of ten after a certain number of iterationsHere we have used dynamic velocity control variant ofPSO in which velocity is scaled down after every specificnumber of generationsThismodification results in improvedperformance as compared to the original algorithm In thispaper the parameters used for PSO algorithm for finding
Frequency
Mag
nitu
de re
spon
se
Magnitude responses of proposed half integrators
Ideal
Gupta-Madhu-JainKrishna half integrator
25
2
15
1
050 02 04 06 08 1 12 14 16 18 2
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
based on PSO optimized DIs
Figure 5 Comparison of magnitude responses of proposed halfintegrators based on 2nd and 3rd order optimized DIs with existing[2 17] models
optimized solutions of 2nd and 3rd order DIs are as followsparticle range = (minus1 1) learning factors (119888
1and 119888
2) = 2
diversity factor = 1 maximum velocity factor = 05 swarmsize = 200 and number of generations = 200 The flowchartof modified PSO has been given in Figure 1
Intel Core 2 Duo CPU T6600 220GHz (InstalledMemory (RAM) of 4GB) has been used for simulationsTransfer functions of worst and average cases of 2nd and 3rdorderDIs have been given inTables 1 and 2 respectively alongwith their corresponding fitness function values (FV)
In this paper when the code was run in C++ for 100iterations the best optimized (with least mean square error)results for 2nd and 3rd order proposed DIs have beenobserved for 200 generations using 200 particles Transferfunctions (TFs) of proposed optimized integrators have beengiven next in pole zero form in (12) and (13) respectively
TF of proposed 2nd order optimized integrator is
119867PSO 2nd opt (119911) = ((08651 (119911 + 01042) (119911 + 06276))
((119911 minus 1) (119911 + 05547)))
(12)
TF of proposed 3rd order optimized integrator is
119867PSO 3rd opt (119911)
= ((08656 (119911 + 01057) (119911 + 02021) (119911 + 0503))
((119911 + 038) (119911 + 02522) (119911 minus 09997)))
(13)
Proposed 2nd and 3rd order DIs optimized here by PSOhave been compared with the recently published integratorsoptimized by different optimization algorithms namely LP
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Optimization
andMaione [32] which proved like boon for the developmentof fractional integratorsdifferentiators
In spite of the efficient and systematic nature of dif-ferent previously mentioned approximations still thereexisted a dissent about performance of these approximationswhen these were later adapted by well-known discretiza-tion schemes for obtaining FOIs Moreover development ofnew approximations requiredmany exhaustivemathematicalapplications to make them perform in finite dimensionalform in proper range of frequencies during discretizationInstead of deriving new methods of rational approximationsauthors have suggested here an alternate solution in whichthe integrator operator (used as 119904-119911 transform) itself shouldbe pushed near boundaries of an optimal solution (near idealresponse) before converting it to fractional system Reasonbehind this is that every minute dispensation in propertiesof this integer order operator will directly propel into itsnoninteger order counterpart The main novelty proposed bythis paper is to put together unique combination of opti-mized integer order operators and an accurate approximationtechnique for configuring their efficient fractional order (fo)operators The current scenario of design methods for deriv-ing optimal operators [33ndash41] is cumulatively proceedingtowards ldquonot so distantrdquo phase of perfect optimization withnegligible errors So another motivating point behind theapproach proposed in this paper is to trespass all disadvan-tages of existing conventional design methods of FOIs byusing refined capabilities of an efficient optimization scheme
The frequency response of ideal integer order digitalintegrators (DIs) is
119867int (120596) = (1
119895120596) where |120596| le 120587 (2)
and the frequency response of ideal integer order digital dif-ferentiators (DDs) is
119867diff (120596) = 119895120596 where |120596| le 120587 (3)
where 119895 = radicminus1 120596 is the angular frequency in radians andldquo119879rdquo is the sampling period
In recent novel advancements different optimizationalgorithms namely Linear Programming (LP) [33 34]Genetic Algorithm (GA) [35ndash37] Simulated Annealing (SA)[38 39] and pole-zero (PZ) optimization [40 41] havebeen used for obtaining more refined integer order differ-integrators Jain et al [35] have recently derived optimalmodels of recursive DIs and DDs by GA algorithm whileUpadhyay [41] has developed these operators by analyzingpoles and zeros of existing recursive DDs by PZ optimizationtechnique for obtaining better results for frequency responsesover wideband of complete frequency spectrum Al-Alaoui[39] has done a credible work in improving design processof DIs and DDs based on Newton-Cotes integration rules byusing linear interpolation and SA
This paper attempts to find the optimal solutions ofDIs by using an efficient modified application of Parti-cle Swarm Optimization (PSO) technique [42ndash44] beforediscretizing them to their fractional mathematical models
Values of mean median and standard deviation in differentrandom iterations have been continuously registered usingPSO for the optimization of 2nd and 3rd order DIs withminimized error fitness function Later these optimized DIsare discretized by CFE of indirect discretization scheme [23]for finding half integrator models of different orders Thesimulation results of magnitude responses phase responsesand relative magnitude errors for the proposed optimizedDIs have been compared with all the recently publishedintegrators optimized by different optimization algorithmsnamely LP [34] GA [35] SA [39] and PZ [41] Frequencyresponses of the proposed FOIs based on 2nd and 3rd orderoptimized DIs have been also compared with their existingfractional models [2 17] in complete frequency range
The paper is organized as follows Section 2 deals withthe brief description of original and modified PSO algo-rithm for optimizing DIs along with its trade-off con-ditions and the resultant 2nd and 3rd order optimizedintegrators thus derived have been presented in this sec-tion Comparisons of proposed and existing DIs are alsogiven in the same section These optimized operators havebeen discretized by indirect discretization using CFE tech-nique of indirect discretization for deriving models of FOIsin Section 3 Section 4 presents the simulation results ofcomparisons of proposed half integrators with responsesof ideal and the existing models Section 5 concludes thepaper
2 PSO Algorithm for Optimizing 2nd and 3rdOrder Integer Order Integrators
21 Outline of Basic Functionality of Original PSO AlgorithmIn conventional PSO algorithm [42ndash44] first a popula-tion of random solutions has been initialized and differentgenerations are updated for searching optimal values Inthis algorithm the potential solutions called particles flythrough the problem space by following the current optimumparticles The position of each particle encodes a possiblesolution to the given problem The velocity of the particle isthe parameter to be added to the current position to find thenewposition of the next generationThe velocity and positionare updated for all particles in every generation unless anoptimal solution is found The particles keep track of itscoordinates in the problem space which are associated withthe best solution (fitness) it has achieved so far These valuesare called pbest and the best element among all the pbestvalues of all the particles is called gbest The velocities areinitialized to zero and after finding the two best values theparticle updates its velocity and positions in accordance withthe following update equations
V (119894) = V (119894 minus 1) + 11988811199031
(119901pbest minus 119901 (119894)) + 11988821199032
(119901gbest minus 119901 (119894))
119901 (119894 + 1) = 119901 (119894) + V (119894)
(4)
where V(119894) is the current particle velocity V(119894 minus 1) is previousparticle velocity 119901(119894) is current particle position 119901(119894 + 1) isthe position of particle in next generation 119901pbest is the best
Journal of Optimization 3
Define the solution space fitness function and population size
Initialise position velocity
For each particle
Evaluate fitness function
Next particle
orNext iteration
Solution is gbest
Update velocityand position
Yes
No
If fitness (x) better than
pbest and gbest
fitness (gbest)gbest = x
No iteration = max no iteration
fitness (gbest) is good enough
minus |H(ej120596)|)2d120596
If fitness (x) better thanfitness (pbest)
pbest = xEint = int
120587
0(1120596
Figure 1 Flowchart of modified PSO algorithm
position the particle has seen 119901gbests is the best position theswarm has seen 119888
1 1198882are constants set to 2 and 119903
1 1199032are the
random numbers between 0 and 1
22 Parameters Used for Optimization and Trade-Offs betweenMultiple Parameters during Convergence of Optimized DigitalIntegrators Amathematical model of transfer function (TF)of optimized digital integrator (DI) can be described ingeneralized form as
119867 (119911) = (119861119872
(119911)119887119872 + sdot sdot sdot + 119861
1(119911)1198871 + 119861
0(119911)1198870
119860119873
(119911)119886119873 + sdot sdot sdot + 119860
1(119911)1198861 + 119860
0(119911)1198860
) (5)
where119860119894and 119861
119895 are arbitrary constants whereas 119886
119894and 119887119895are
real numbers with (119894 = 0 1 119873) and (119895 = 0 1 119872)
Transfer functions (TFs) of 2nd and 3rd order DIs whichhave been optimized using PSO in this paper are given in (6)and (7) respectively
119867PSO 2nd opt (119911) = (11986121199112 + 119861
1119911 + 1198610
11986021199112 + 119860
1119911 + 119860
0
) (6)
119867PSO 3rd opt (119911) = (1198613(119911)3
+ 1198612(119911)2
+ 1198611119911 + 1198610
1198603(119911)3
+ 1198602(119911)2
+ 1198601119911 + 119860
0
) (7)
The coefficients of the previously mentioned TFs are gener-ated by random iterations of PSO algorithm as it takes realnumbers as particles
4 Journal of Optimization
Table 1 Transfer functions and fitness values of worst and average cases of 2nd order PSO optimized DIs
TFs of worst cases of 2nd order DIs along with their fitnessfunction values (FV)
TFs of average cases of 2nd order DIs along with their fitnessfunction values (FV)
Transfer function FV Transfer function FV
1198671(119911) = (
03733241199112 + 0427523119911 + 0397575
1199112 minus 0042561119911 minus 0999916) 0643052 119867
1(119911) = (
0228961199112 minus 0825829119911 minus 0502494
1199112 minus 0776284119911 minus 0181729) 0382107
1198672(119911) = (
minus09292511199112 + 0381076119911 minus 075834
1199112 minus 0725805119911 minus 0306558) 390016 119867
2(119911) = (
minus03364781199112 minus 082224119911 + 0101197
1199112 minus 0818946119911 minus 0147828) 0407093
1198673(119911) = (
05416991199112 + 0241831119911 + 0506633
1199112 minus 00552516119911 minus 1) 473234 119867
3(119911) = (
008498351199112 + 0752164119911 + 0756218
1199112 minus 0167219119911 minus 0882869) 050102
1198674(119911) = (
minus03545831199112 minus 0833923119911 minus 0813869
1199112 minus 00637125119911 minus 1) 589578 119867
4(119911) = (
minus0397291199112 minus 0795585119911 minus 0130415
1199112 minus 0722962119911 minus 0318656) 0618717
1198675(119911) = (
03538521199112 + 0203374119911 + 0466221
1199112 minus 00406162119911 minus 1) 18799 119867
5(119911) = (
02843821199112 + 0949776119911 + 0363676
1199112 minus 0401952119911 minus 0648204) 0737534
Magnitude response of the optimized DI is
119867mag = |119867 (119908)| =abs (numerator)abs (denominator)
(8)
Phase response of the optimized DI is
119867phase = arg |119867 (119908)| (9)
Relative magnitude error (dB) or RME value is given by
RME = 20 log10
100381610038161003816100381610038161003816100381610038161003816
119867ideal (120596) minus 119867approx (120596)
119867ideal (120596)
100381610038161003816100381610038161003816100381610038161003816
= 20 log10
1003816100381610038161003816100381610038161003816100381610038161003816
(1119895120596) minus 119867approx (120596)
(1119895120596)
1003816100381610038161003816100381610038161003816100381610038161003816
(10)
where119867ideal(120596) is themagnitude of ideal value of the operatorand 119867approx(120596) is the value of the digital integrator that isapproximated
In the recently published literature on digital and frac-tional operators most of the researchers have focused onformulating the fitness function for optimization techniquebased on only single parameter [38ndash41] that is magnitudeerror because practically if phase and magnitude are opti-mized for a multiparameter problem the solution is alwaysstuck in a trade-off due to their contradictory nature Thereason behind this is that when the optimal solution startsconverging towards the best results of one parameter (mag-nitude response) it simultaneously degrades optimality ofsecond parameter (phase responses) after certain number ofiterations Solution for such trade-offs is to choose domi-nating parameter in problem under consideration In thispaper the design strategy to find optimized DIs is kept moreconcerned for minimum magnitude response error whichis comparatively more significant characteristic The fitnessfunction used to approximate a digital integrator is taken asthe mean square error of the transfer function and it is givenas
119864int = int120587
0
(1
120596minus
10038161003816100381610038161003816119867 (119890119895120596
)10038161003816100381610038161003816)2
119889120596 (11)
1 15 2 25Frequency
Mag
nitu
de re
spon
se
Comparison of magnitude responses of proposed DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
1
09
08
07
06
05
04
03
02
with existing optimized DIs
Figure 2 Comparison ofmagnitude responses of proposed 2nd and3rd order optimized DIs with existing [26 34 39 41] optimizedmodels
Aim of PSO optimization is to search the best optimalsolution (TF of DI) with minimized mean square error
23 Modified PSO Algorithm for Derivation of OptimizedDIs In this paper a slight modification in the originalPSO algorithm has been introduced in velocity taking intoaccount the given problem of deriving the best feasibleoptimized DIs with very small magnitude response errorswithout worsening their corresponding phase responses Inmodified PSO used in this paper the maximum velocitywhich constraints the movement of swarms to the range is
Journal of Optimization 5
0 05 1 15 2 25 3
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Comparison of RME values of proposed DIs
minus20
minus40
minus60
minus80
minus100
minus120
Al-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)Upadhyay (pole-zero optim)
Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
with existing optimized DIs
Figure 3 Comparison of RMEvalues of proposed 2nd and 3rd orderoptimized DIs with existing [26 34 39 41] optimized models
0
50
100
150
200
Frequency
Phas
e (de
g)
Comparison of phase responses of of proposed DIs
minus50
minus100
minus150
minus2000 05 1 15 2 25 3
with existing optimized DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
Figure 4 Comparison of phase responses of proposed 2nd and 3rdorder optimized DIs with existing [26 34 39 41] optimized models
divided by a factor of ten after a certain number of iterationsHere we have used dynamic velocity control variant ofPSO in which velocity is scaled down after every specificnumber of generationsThismodification results in improvedperformance as compared to the original algorithm In thispaper the parameters used for PSO algorithm for finding
Frequency
Mag
nitu
de re
spon
se
Magnitude responses of proposed half integrators
Ideal
Gupta-Madhu-JainKrishna half integrator
25
2
15
1
050 02 04 06 08 1 12 14 16 18 2
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
based on PSO optimized DIs
Figure 5 Comparison of magnitude responses of proposed halfintegrators based on 2nd and 3rd order optimized DIs with existing[2 17] models
optimized solutions of 2nd and 3rd order DIs are as followsparticle range = (minus1 1) learning factors (119888
1and 119888
2) = 2
diversity factor = 1 maximum velocity factor = 05 swarmsize = 200 and number of generations = 200 The flowchartof modified PSO has been given in Figure 1
Intel Core 2 Duo CPU T6600 220GHz (InstalledMemory (RAM) of 4GB) has been used for simulationsTransfer functions of worst and average cases of 2nd and 3rdorderDIs have been given inTables 1 and 2 respectively alongwith their corresponding fitness function values (FV)
In this paper when the code was run in C++ for 100iterations the best optimized (with least mean square error)results for 2nd and 3rd order proposed DIs have beenobserved for 200 generations using 200 particles Transferfunctions (TFs) of proposed optimized integrators have beengiven next in pole zero form in (12) and (13) respectively
TF of proposed 2nd order optimized integrator is
119867PSO 2nd opt (119911) = ((08651 (119911 + 01042) (119911 + 06276))
((119911 minus 1) (119911 + 05547)))
(12)
TF of proposed 3rd order optimized integrator is
119867PSO 3rd opt (119911)
= ((08656 (119911 + 01057) (119911 + 02021) (119911 + 0503))
((119911 + 038) (119911 + 02522) (119911 minus 09997)))
(13)
Proposed 2nd and 3rd order DIs optimized here by PSOhave been compared with the recently published integratorsoptimized by different optimization algorithms namely LP
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Optimization 3
Define the solution space fitness function and population size
Initialise position velocity
For each particle
Evaluate fitness function
Next particle
orNext iteration
Solution is gbest
Update velocityand position
Yes
No
If fitness (x) better than
pbest and gbest
fitness (gbest)gbest = x
No iteration = max no iteration
fitness (gbest) is good enough
minus |H(ej120596)|)2d120596
If fitness (x) better thanfitness (pbest)
pbest = xEint = int
120587
0(1120596
Figure 1 Flowchart of modified PSO algorithm
position the particle has seen 119901gbests is the best position theswarm has seen 119888
1 1198882are constants set to 2 and 119903
1 1199032are the
random numbers between 0 and 1
22 Parameters Used for Optimization and Trade-Offs betweenMultiple Parameters during Convergence of Optimized DigitalIntegrators Amathematical model of transfer function (TF)of optimized digital integrator (DI) can be described ingeneralized form as
119867 (119911) = (119861119872
(119911)119887119872 + sdot sdot sdot + 119861
1(119911)1198871 + 119861
0(119911)1198870
119860119873
(119911)119886119873 + sdot sdot sdot + 119860
1(119911)1198861 + 119860
0(119911)1198860
) (5)
where119860119894and 119861
119895 are arbitrary constants whereas 119886
119894and 119887119895are
real numbers with (119894 = 0 1 119873) and (119895 = 0 1 119872)
Transfer functions (TFs) of 2nd and 3rd order DIs whichhave been optimized using PSO in this paper are given in (6)and (7) respectively
119867PSO 2nd opt (119911) = (11986121199112 + 119861
1119911 + 1198610
11986021199112 + 119860
1119911 + 119860
0
) (6)
119867PSO 3rd opt (119911) = (1198613(119911)3
+ 1198612(119911)2
+ 1198611119911 + 1198610
1198603(119911)3
+ 1198602(119911)2
+ 1198601119911 + 119860
0
) (7)
The coefficients of the previously mentioned TFs are gener-ated by random iterations of PSO algorithm as it takes realnumbers as particles
4 Journal of Optimization
Table 1 Transfer functions and fitness values of worst and average cases of 2nd order PSO optimized DIs
TFs of worst cases of 2nd order DIs along with their fitnessfunction values (FV)
TFs of average cases of 2nd order DIs along with their fitnessfunction values (FV)
Transfer function FV Transfer function FV
1198671(119911) = (
03733241199112 + 0427523119911 + 0397575
1199112 minus 0042561119911 minus 0999916) 0643052 119867
1(119911) = (
0228961199112 minus 0825829119911 minus 0502494
1199112 minus 0776284119911 minus 0181729) 0382107
1198672(119911) = (
minus09292511199112 + 0381076119911 minus 075834
1199112 minus 0725805119911 minus 0306558) 390016 119867
2(119911) = (
minus03364781199112 minus 082224119911 + 0101197
1199112 minus 0818946119911 minus 0147828) 0407093
1198673(119911) = (
05416991199112 + 0241831119911 + 0506633
1199112 minus 00552516119911 minus 1) 473234 119867
3(119911) = (
008498351199112 + 0752164119911 + 0756218
1199112 minus 0167219119911 minus 0882869) 050102
1198674(119911) = (
minus03545831199112 minus 0833923119911 minus 0813869
1199112 minus 00637125119911 minus 1) 589578 119867
4(119911) = (
minus0397291199112 minus 0795585119911 minus 0130415
1199112 minus 0722962119911 minus 0318656) 0618717
1198675(119911) = (
03538521199112 + 0203374119911 + 0466221
1199112 minus 00406162119911 minus 1) 18799 119867
5(119911) = (
02843821199112 + 0949776119911 + 0363676
1199112 minus 0401952119911 minus 0648204) 0737534
Magnitude response of the optimized DI is
119867mag = |119867 (119908)| =abs (numerator)abs (denominator)
(8)
Phase response of the optimized DI is
119867phase = arg |119867 (119908)| (9)
Relative magnitude error (dB) or RME value is given by
RME = 20 log10
100381610038161003816100381610038161003816100381610038161003816
119867ideal (120596) minus 119867approx (120596)
119867ideal (120596)
100381610038161003816100381610038161003816100381610038161003816
= 20 log10
1003816100381610038161003816100381610038161003816100381610038161003816
(1119895120596) minus 119867approx (120596)
(1119895120596)
1003816100381610038161003816100381610038161003816100381610038161003816
(10)
where119867ideal(120596) is themagnitude of ideal value of the operatorand 119867approx(120596) is the value of the digital integrator that isapproximated
In the recently published literature on digital and frac-tional operators most of the researchers have focused onformulating the fitness function for optimization techniquebased on only single parameter [38ndash41] that is magnitudeerror because practically if phase and magnitude are opti-mized for a multiparameter problem the solution is alwaysstuck in a trade-off due to their contradictory nature Thereason behind this is that when the optimal solution startsconverging towards the best results of one parameter (mag-nitude response) it simultaneously degrades optimality ofsecond parameter (phase responses) after certain number ofiterations Solution for such trade-offs is to choose domi-nating parameter in problem under consideration In thispaper the design strategy to find optimized DIs is kept moreconcerned for minimum magnitude response error whichis comparatively more significant characteristic The fitnessfunction used to approximate a digital integrator is taken asthe mean square error of the transfer function and it is givenas
119864int = int120587
0
(1
120596minus
10038161003816100381610038161003816119867 (119890119895120596
)10038161003816100381610038161003816)2
119889120596 (11)
1 15 2 25Frequency
Mag
nitu
de re
spon
se
Comparison of magnitude responses of proposed DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
1
09
08
07
06
05
04
03
02
with existing optimized DIs
Figure 2 Comparison ofmagnitude responses of proposed 2nd and3rd order optimized DIs with existing [26 34 39 41] optimizedmodels
Aim of PSO optimization is to search the best optimalsolution (TF of DI) with minimized mean square error
23 Modified PSO Algorithm for Derivation of OptimizedDIs In this paper a slight modification in the originalPSO algorithm has been introduced in velocity taking intoaccount the given problem of deriving the best feasibleoptimized DIs with very small magnitude response errorswithout worsening their corresponding phase responses Inmodified PSO used in this paper the maximum velocitywhich constraints the movement of swarms to the range is
Journal of Optimization 5
0 05 1 15 2 25 3
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Comparison of RME values of proposed DIs
minus20
minus40
minus60
minus80
minus100
minus120
Al-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)Upadhyay (pole-zero optim)
Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
with existing optimized DIs
Figure 3 Comparison of RMEvalues of proposed 2nd and 3rd orderoptimized DIs with existing [26 34 39 41] optimized models
0
50
100
150
200
Frequency
Phas
e (de
g)
Comparison of phase responses of of proposed DIs
minus50
minus100
minus150
minus2000 05 1 15 2 25 3
with existing optimized DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
Figure 4 Comparison of phase responses of proposed 2nd and 3rdorder optimized DIs with existing [26 34 39 41] optimized models
divided by a factor of ten after a certain number of iterationsHere we have used dynamic velocity control variant ofPSO in which velocity is scaled down after every specificnumber of generationsThismodification results in improvedperformance as compared to the original algorithm In thispaper the parameters used for PSO algorithm for finding
Frequency
Mag
nitu
de re
spon
se
Magnitude responses of proposed half integrators
Ideal
Gupta-Madhu-JainKrishna half integrator
25
2
15
1
050 02 04 06 08 1 12 14 16 18 2
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
based on PSO optimized DIs
Figure 5 Comparison of magnitude responses of proposed halfintegrators based on 2nd and 3rd order optimized DIs with existing[2 17] models
optimized solutions of 2nd and 3rd order DIs are as followsparticle range = (minus1 1) learning factors (119888
1and 119888
2) = 2
diversity factor = 1 maximum velocity factor = 05 swarmsize = 200 and number of generations = 200 The flowchartof modified PSO has been given in Figure 1
Intel Core 2 Duo CPU T6600 220GHz (InstalledMemory (RAM) of 4GB) has been used for simulationsTransfer functions of worst and average cases of 2nd and 3rdorderDIs have been given inTables 1 and 2 respectively alongwith their corresponding fitness function values (FV)
In this paper when the code was run in C++ for 100iterations the best optimized (with least mean square error)results for 2nd and 3rd order proposed DIs have beenobserved for 200 generations using 200 particles Transferfunctions (TFs) of proposed optimized integrators have beengiven next in pole zero form in (12) and (13) respectively
TF of proposed 2nd order optimized integrator is
119867PSO 2nd opt (119911) = ((08651 (119911 + 01042) (119911 + 06276))
((119911 minus 1) (119911 + 05547)))
(12)
TF of proposed 3rd order optimized integrator is
119867PSO 3rd opt (119911)
= ((08656 (119911 + 01057) (119911 + 02021) (119911 + 0503))
((119911 + 038) (119911 + 02522) (119911 minus 09997)))
(13)
Proposed 2nd and 3rd order DIs optimized here by PSOhave been compared with the recently published integratorsoptimized by different optimization algorithms namely LP
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Journal of Optimization
Table 1 Transfer functions and fitness values of worst and average cases of 2nd order PSO optimized DIs
TFs of worst cases of 2nd order DIs along with their fitnessfunction values (FV)
TFs of average cases of 2nd order DIs along with their fitnessfunction values (FV)
Transfer function FV Transfer function FV
1198671(119911) = (
03733241199112 + 0427523119911 + 0397575
1199112 minus 0042561119911 minus 0999916) 0643052 119867
1(119911) = (
0228961199112 minus 0825829119911 minus 0502494
1199112 minus 0776284119911 minus 0181729) 0382107
1198672(119911) = (
minus09292511199112 + 0381076119911 minus 075834
1199112 minus 0725805119911 minus 0306558) 390016 119867
2(119911) = (
minus03364781199112 minus 082224119911 + 0101197
1199112 minus 0818946119911 minus 0147828) 0407093
1198673(119911) = (
05416991199112 + 0241831119911 + 0506633
1199112 minus 00552516119911 minus 1) 473234 119867
3(119911) = (
008498351199112 + 0752164119911 + 0756218
1199112 minus 0167219119911 minus 0882869) 050102
1198674(119911) = (
minus03545831199112 minus 0833923119911 minus 0813869
1199112 minus 00637125119911 minus 1) 589578 119867
4(119911) = (
minus0397291199112 minus 0795585119911 minus 0130415
1199112 minus 0722962119911 minus 0318656) 0618717
1198675(119911) = (
03538521199112 + 0203374119911 + 0466221
1199112 minus 00406162119911 minus 1) 18799 119867
5(119911) = (
02843821199112 + 0949776119911 + 0363676
1199112 minus 0401952119911 minus 0648204) 0737534
Magnitude response of the optimized DI is
119867mag = |119867 (119908)| =abs (numerator)abs (denominator)
(8)
Phase response of the optimized DI is
119867phase = arg |119867 (119908)| (9)
Relative magnitude error (dB) or RME value is given by
RME = 20 log10
100381610038161003816100381610038161003816100381610038161003816
119867ideal (120596) minus 119867approx (120596)
119867ideal (120596)
100381610038161003816100381610038161003816100381610038161003816
= 20 log10
1003816100381610038161003816100381610038161003816100381610038161003816
(1119895120596) minus 119867approx (120596)
(1119895120596)
1003816100381610038161003816100381610038161003816100381610038161003816
(10)
where119867ideal(120596) is themagnitude of ideal value of the operatorand 119867approx(120596) is the value of the digital integrator that isapproximated
In the recently published literature on digital and frac-tional operators most of the researchers have focused onformulating the fitness function for optimization techniquebased on only single parameter [38ndash41] that is magnitudeerror because practically if phase and magnitude are opti-mized for a multiparameter problem the solution is alwaysstuck in a trade-off due to their contradictory nature Thereason behind this is that when the optimal solution startsconverging towards the best results of one parameter (mag-nitude response) it simultaneously degrades optimality ofsecond parameter (phase responses) after certain number ofiterations Solution for such trade-offs is to choose domi-nating parameter in problem under consideration In thispaper the design strategy to find optimized DIs is kept moreconcerned for minimum magnitude response error whichis comparatively more significant characteristic The fitnessfunction used to approximate a digital integrator is taken asthe mean square error of the transfer function and it is givenas
119864int = int120587
0
(1
120596minus
10038161003816100381610038161003816119867 (119890119895120596
)10038161003816100381610038161003816)2
119889120596 (11)
1 15 2 25Frequency
Mag
nitu
de re
spon
se
Comparison of magnitude responses of proposed DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
1
09
08
07
06
05
04
03
02
with existing optimized DIs
Figure 2 Comparison ofmagnitude responses of proposed 2nd and3rd order optimized DIs with existing [26 34 39 41] optimizedmodels
Aim of PSO optimization is to search the best optimalsolution (TF of DI) with minimized mean square error
23 Modified PSO Algorithm for Derivation of OptimizedDIs In this paper a slight modification in the originalPSO algorithm has been introduced in velocity taking intoaccount the given problem of deriving the best feasibleoptimized DIs with very small magnitude response errorswithout worsening their corresponding phase responses Inmodified PSO used in this paper the maximum velocitywhich constraints the movement of swarms to the range is
Journal of Optimization 5
0 05 1 15 2 25 3
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Comparison of RME values of proposed DIs
minus20
minus40
minus60
minus80
minus100
minus120
Al-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)Upadhyay (pole-zero optim)
Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
with existing optimized DIs
Figure 3 Comparison of RMEvalues of proposed 2nd and 3rd orderoptimized DIs with existing [26 34 39 41] optimized models
0
50
100
150
200
Frequency
Phas
e (de
g)
Comparison of phase responses of of proposed DIs
minus50
minus100
minus150
minus2000 05 1 15 2 25 3
with existing optimized DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
Figure 4 Comparison of phase responses of proposed 2nd and 3rdorder optimized DIs with existing [26 34 39 41] optimized models
divided by a factor of ten after a certain number of iterationsHere we have used dynamic velocity control variant ofPSO in which velocity is scaled down after every specificnumber of generationsThismodification results in improvedperformance as compared to the original algorithm In thispaper the parameters used for PSO algorithm for finding
Frequency
Mag
nitu
de re
spon
se
Magnitude responses of proposed half integrators
Ideal
Gupta-Madhu-JainKrishna half integrator
25
2
15
1
050 02 04 06 08 1 12 14 16 18 2
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
based on PSO optimized DIs
Figure 5 Comparison of magnitude responses of proposed halfintegrators based on 2nd and 3rd order optimized DIs with existing[2 17] models
optimized solutions of 2nd and 3rd order DIs are as followsparticle range = (minus1 1) learning factors (119888
1and 119888
2) = 2
diversity factor = 1 maximum velocity factor = 05 swarmsize = 200 and number of generations = 200 The flowchartof modified PSO has been given in Figure 1
Intel Core 2 Duo CPU T6600 220GHz (InstalledMemory (RAM) of 4GB) has been used for simulationsTransfer functions of worst and average cases of 2nd and 3rdorderDIs have been given inTables 1 and 2 respectively alongwith their corresponding fitness function values (FV)
In this paper when the code was run in C++ for 100iterations the best optimized (with least mean square error)results for 2nd and 3rd order proposed DIs have beenobserved for 200 generations using 200 particles Transferfunctions (TFs) of proposed optimized integrators have beengiven next in pole zero form in (12) and (13) respectively
TF of proposed 2nd order optimized integrator is
119867PSO 2nd opt (119911) = ((08651 (119911 + 01042) (119911 + 06276))
((119911 minus 1) (119911 + 05547)))
(12)
TF of proposed 3rd order optimized integrator is
119867PSO 3rd opt (119911)
= ((08656 (119911 + 01057) (119911 + 02021) (119911 + 0503))
((119911 + 038) (119911 + 02522) (119911 minus 09997)))
(13)
Proposed 2nd and 3rd order DIs optimized here by PSOhave been compared with the recently published integratorsoptimized by different optimization algorithms namely LP
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Optimization 5
0 05 1 15 2 25 3
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Comparison of RME values of proposed DIs
minus20
minus40
minus60
minus80
minus100
minus120
Al-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)Upadhyay (pole-zero optim)
Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
with existing optimized DIs
Figure 3 Comparison of RMEvalues of proposed 2nd and 3rd orderoptimized DIs with existing [26 34 39 41] optimized models
0
50
100
150
200
Frequency
Phas
e (de
g)
Comparison of phase responses of of proposed DIs
minus50
minus100
minus150
minus2000 05 1 15 2 25 3
with existing optimized DIs
IdealAl-Alaoui 3-segment (SA)Al-Alaoui 4-segment (SA)Jain-Gupta-Jain (GA)
Upadhyay (pole-zero optim)Gupta-Jain-Kumar (LP)Proposed 2nd orderProposed 3rd order
Figure 4 Comparison of phase responses of proposed 2nd and 3rdorder optimized DIs with existing [26 34 39 41] optimized models
divided by a factor of ten after a certain number of iterationsHere we have used dynamic velocity control variant ofPSO in which velocity is scaled down after every specificnumber of generationsThismodification results in improvedperformance as compared to the original algorithm In thispaper the parameters used for PSO algorithm for finding
Frequency
Mag
nitu
de re
spon
se
Magnitude responses of proposed half integrators
Ideal
Gupta-Madhu-JainKrishna half integrator
25
2
15
1
050 02 04 06 08 1 12 14 16 18 2
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
based on PSO optimized DIs
Figure 5 Comparison of magnitude responses of proposed halfintegrators based on 2nd and 3rd order optimized DIs with existing[2 17] models
optimized solutions of 2nd and 3rd order DIs are as followsparticle range = (minus1 1) learning factors (119888
1and 119888
2) = 2
diversity factor = 1 maximum velocity factor = 05 swarmsize = 200 and number of generations = 200 The flowchartof modified PSO has been given in Figure 1
Intel Core 2 Duo CPU T6600 220GHz (InstalledMemory (RAM) of 4GB) has been used for simulationsTransfer functions of worst and average cases of 2nd and 3rdorderDIs have been given inTables 1 and 2 respectively alongwith their corresponding fitness function values (FV)
In this paper when the code was run in C++ for 100iterations the best optimized (with least mean square error)results for 2nd and 3rd order proposed DIs have beenobserved for 200 generations using 200 particles Transferfunctions (TFs) of proposed optimized integrators have beengiven next in pole zero form in (12) and (13) respectively
TF of proposed 2nd order optimized integrator is
119867PSO 2nd opt (119911) = ((08651 (119911 + 01042) (119911 + 06276))
((119911 minus 1) (119911 + 05547)))
(12)
TF of proposed 3rd order optimized integrator is
119867PSO 3rd opt (119911)
= ((08656 (119911 + 01057) (119911 + 02021) (119911 + 0503))
((119911 + 038) (119911 + 02522) (119911 minus 09997)))
(13)
Proposed 2nd and 3rd order DIs optimized here by PSOhave been compared with the recently published integratorsoptimized by different optimization algorithms namely LP
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Journal of Optimization
0
20
40
Frequency
Rel
ativ
e mag
nitu
de er
ror (
dB)
Relative magnitude errors (dB) of proposed half integrators
minus20
minus40
minus60
minus80
minus100
minus1200 05 1 15 2 25 3
based on PSO optimized DIs
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)H3rd fo 3(z)
H3rd fo 6(z)
Figure 6 Comparison of RME values of proposed half integratorsbased on 2nd and 3rd order optimized DIs with existing [2 17]models
[34] GA [35] SA [39] and PZ [41] for validating efficientperformance of proposed PSO Comparisons of bode dia-grams (ie magnitude versus the frequency (119908) and phase(degrees) versus frequency (119908)) and RME responses (in dB)have been presented in Figures 2 3 and 4 respectively
Results for performance values (RME) for each opti-mization method have been compared in Table 3 It canbe observed from Figure 3 that the proposed PSO methodclearly outperforms existing SA [39] and LP [34] techniquesand gives comparable results for PZ [41] and GA [35] meth-ods The 2nd order DI namely 119867PSO 2nd opt(119911) given in (12)outperforms Al-Alaoui 3-segment and 4-segment optimizedDIs in complete spectrum 119867PSO 2nd opt(119911) excels Jain-Gupta-Jain DI [35] in frequency range over 0 le 120596 le 05120587 radiansand 258 le 120596 le 3120587 radians PSO optimized proposed 3rdorder DI namely 119867PSO 3rd opt(119911) clearly outperforms theexisting Al-Alaoui optimized 3-segment 4-segment [39] andUpadhyay [41] DIs in almost complete range with RME of le
minus41 dB 119867PSO 3rd opt(119911) given in (13) gives comparable resultswith GA optimized [35] operator Proposed optimized DIsare observed to satisfy stability criterion and show linearlydecreasing phase responses (see Figure 4)
Proposed DIs optimized by PSO clearly excel operatorsoptimized by SA [39] LP [34] and PZ [41] techniques butlie in close vicinity of the GA [35] responses So a statisticaltest has been done for comparing performances of PSOand GA for validating superiority of the proposed modifiedalgorithm The products of number of particles and numberof generations (119873
119901lowast 119873119892) have been varied from (50 lowast 50) to
(100 lowast 100) and then to (200 lowast 200) for different iterationsand values of mean median and standard deviation have
Pha
se (d
eg)
Phase responses of proposed half integrators
10
0
minus10
minus20
minus30
minus40
minus50
based on PSO optimized DIs
Frequency0 05 1 15 2 25 3
Ideal
Gupta-Madhu-JainKrishna half integrator
H2nd fo 2(z)H2nd fo 4(z)H2nd fo 6(z)
H3rd fo 3(z)H3rd fo 6(z)
Figure 7 Comparison of phase responses of proposed half integra-tors based on 2nd and 3rd order optimized DIs with existing [2 17]models
been registered for all the worst average and best results of2nd and 3rd order optimized DIs (see Table 4) From thistable it is clear that PSO technique performs better than GAmethod and also it was noticed that with the increase invalues of 119873
119901 119873119892 and number of iterations the response of
resultant DIs comes closer to the ideal response as meanmedian and standard deviation values go on decreasing withefficientminimization of error function In this paperwe haveconsidered values of root mean square error
3 Discretization ofProposed 2nd and 3rd Order OptimizedDIs for Half Integrator Models
Theoptimized integer order DIs derived in Section 2 are usedas 119904-to-119911 transformations for CFE based indirect discretiza-tion [2 23 24] scheme Formula given by Khovanskii in [29]has been used for different rational approximations of FOIsThat is
(119904)plusmn120572
=1
1minus
120572119904
1+(1+120572) 119904minus
(1 + 120572) 119904 (1 + 119904)
2+(3+120572) 119904minus
2 (2 + 120572) 119904 (1 + 119904)
3 + (5+120572) 119904minus
times3 (3 + 120572) 119904 (1 + 119904)
4 + (9 + 120572) 119904minus
4 (4 + 120572) 119904 (1 + 119904)
5 + (9 + 120572) 119904minus
times5 (5 + 120572) 119904 (1 + 119904)
6 + (11 + 120572) 119904minus
6 (6 + 120572) 119904 (1 + 119904)
7 + (13 + 120572) 119904minus
(14)
Infinite series of (14) can be terminated up to 2 4and 6 terms and CFE based indirect discretization used in
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Optimization 7
Table2Transfe
rfun
ctions
andfitnessvalues
ofworstandaveragec
ases
of3rdorderP
SOop
timized
DIs
TFso
fworstcaseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
TFso
faverage
caseso
f3rd
orderD
Isalon
gwith
theirfi
tnessfun
ctionvalues
(FV)
Transfe
rfun
ction
FV
Transfe
rfun
ction
FV
1198671(119911
)=
(01
423941199113
+04
285111199112
minus03
06611119911
+02
17758
1199113
minus09
031351199112
minus06
74295119911
+05
68774
)066
6364
1198671(119911
)=
(minus
06
56851199113
minus08
062451199112
minus08
20988119911
minus01
9093
1199113
+01
005341199112
minus04
73273119911
minus05
50301
)0548217
1198672(119911
)=
(minus
07
387211199113
minus04
255611199112
minus07
87252119911
+05
88187
1199113
minus00
8455311199112
minus07
14103119911
minus02
4424
)238383
1198672(119911
)=
(minus
01
711661199113
+07
950271199112
+03
39133119911
+04
81229
1199113
minus07
403291199112
minus00
821587119911
minus01
30453
)0651617
1198673(119911
)=
(05
248521199113
+09
588131199112
+05
85323119911
+07
79817
1199113
minus04
627791199112
minus03
80859119911
minus02
47247
)391385
1198673(119911
)=
(05
229891199113
+01
636031199112
minus00
262705119911
+04
83852
1199113
minus09
798611199112
minus01
40418119911
+00
831846
)0673745
1198674(119911
)=
(09
790991199113
minus08
529211199112
+09
60434119911
+08
75076
1199113
minus08
270921199112
minus02
46012119911
+01
35716
)406169
1198674(119911
)=
(02
258041199113
+07
800811199112
+07
58662119911
minus00
994224
1199113
minus02
373481199112
minus05
75985119911
minus01
3299
)0713815
1198675(119911
)=
(00
5946561199113
+07
000181199112
+03
46537119911
+03
14347
1199113
minus07
918671199112
minus05
73836119887119911
+04
10591
)468934
1198675(119911
)=
(01
238551199113
+04
917661199112
+07
27444119911
+05
68523
1199113
minus03
196421199112
minus02
87981119911
minus03
31056
)09760
67
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Journal of Optimization
Table 3 Comparisons of RME values of the proposed PSO optimized DIs with other integer order existing integrators optimized by differentoptimization techniques
120596 (120587 rad)Al-Alaoui3-segment
(SA)
Al-Alaoui4-segment
(SA)
Jain-Gupta-Jain(GA)
Upadhyay(P-Z)
Gupta-Jain-Kumar(LP)
Proposed2nd orderDI (PSO)
Proposed3rd orderDI (PSO)
025 minus7203 minus6018 minus5446 minus4831 minus7106 minus7203 minus638205 minus4915 minus4121 minus5947 minus5165 minus5989 minus5968 minus7304075 minus3725 minus3402 minus8004 minus6171 minus6391 minus5327 minus714510 minus3029 minus3034 minus5728 minus5728 minus4994 minus4988 minus6463125 minus2272 minus3296 minus5384 minus4871 minus3933 minus4871 minus682175 minus1164 minus3505 minus5385 minus4791 minus3146 minus6262 minus54220 minus824 minus1911 minus6053 minus5289 minus3143 minus4998 minus4994225 minus692 minus1599 minus5384 minus615 minus345 minus4277 minus5366275 minus1819 minus1973 minus3754 minus3776 minus4588 minus5555 minus406130 minus417 minus1811 minus3552 minus3644 minus3014 minus4134 minus4501
Table 4 Statistical test for comparison of proposed PSO and GA techniques for approximations of 2nd and 3rd order digital integrators
Genetic algorithm Modified PSO algorithmApproximating 2nd order digital integrator for comparing PSO and GA
119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915405 0230096 00874599 0614370 0201256 00936231Median 0482815 0194402 0075695 0491575 0114377 00488172Standard deviation 0737526 0131617 00400587 043579 0137471 0042671
Approximating 3rd order digital integrator for comparing PSO and GA119873119901
lowast 119873119892
50 lowast 50 100 lowast 100 200 lowast 200 50 lowast 50 100 lowast 100 200 lowast 200
Mean 0915406 0230096 00874599 0633554 0221398 00969881Median 01478565 0069402 002937085 01801123 01207528 00861992Standard deviation 0737526 0131617 00400587 0491451 0152073 00530145
[24] has been adopted for deriving half integrators fitted incontinuous-time domain (see Table 5) Also 2nd order opti-mized DI resulted in 2nd 4th and 6th order FOIs whereas3rd order optimized DI when substituted for different termsresulted in 3rd and 6th order half integrators In the TFs ofderived proposed half integrators given in Table 5 order ofoptimized DIs and order of FOIs are used as subscripts ofsymbol ldquo119867rdquo
4 Comparisons of Proposed FOIs Based on2nd and 3rd Order Optimized DIs with theExisting Models
Simulation results of comparison of magnitude responsesRME and phase responses of half integrators (see Table 5)based on different 2nd and 3rd order optimized DIs (given in(12) and (13)) with the existing half integrators given in [2 17]and ideal half integrator have been presented in Figures 5ndash7
Comparison of RME values of proposed and existingFOIs has been presented in Table 6 It is observed fromFigure 6 that the proposed 2nd and 3rd order FOIs namely1198672nd fo 2(119911) and 119867
3rd fo 3(119911) only lag in performance butstill these outperform the existing [2 17] models in higherfrequency spectrumThe6th order operator119867
2nd fo 6(119911) is the
best among the proposed FOIs and it clearly outperforms allthe existing models [2 17] in complete frequency spectrumwith RME of leminus40 dB Proposed 119867
2nd fo 2(119911) and 1198673rd fo 3(119911)
outperform Gupta-Madhu-Jain half integrator [17] in fre-quency range over 088 le 120596 le 3120587 radians 119867
2nd fo 4(119911)1198672nd fo 6(119911) and 119867
3rd fo 6(119911) half integrators excel Krishnahalf integrator [2] in almost complete range All the proposedFOIs are observed to be stable and follow linear phase curvein almost complete frequency spectrum (see Figure 7)
5 Conclusions
This paper attempts to focus on a simple yet unaddressedfact which till date has remained in the backdrop as theoptimization of an integer order operator before convertingit into its fractional the order (fo) counterparts ProposedFOIs validate that properties of an efficiently optimized DIpass directly into its fractional domain when it is used as 119904-to-119911 transformation during indirect discretization Here 2ndand 3rd order integer order integrators have been exploredby PSO algorithm for finding their optimal solutions for aminimized error fitness function The simulation results ofinteger order operators have revealed the effectiveness ofthe proposed modified PSO algorithm because of their low
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Optimization 9
Table 5 Mathematical models of proposed half integrators based on 2nd and 3rd order PSO optimized DIs
Mathematical models of half integrators based on 2nd and 3rd order PSO optimized DIsName of PSO optimized DIs Transfer functions of FOIs derived by indirect discretization of CFE
119867PSO 2nd opt(119911)
1198672nd-fo-2(119911) = (09302 (119911 minus 01825) (119911 + 05869)
(119911 minus 07422) (119911 + 05604))
1198672nd-fo-4(119911) = (093011 (119911 minus 0003372) (119911 + 05633) (119911 + 06071) (119911 minus 06396)
(119911 minus 09034) (119911 minus 0293) (119911 + 05788) (119911 + 05566))
1198672nd-fo-6(119911) = (
(093011 (119911 minus 08073) (119911 minus 03427) (119911 + 0616)
(119911 + 05758) (119911 + 05587) (119911 + 004899))
((119911 + 05936) (119911 + 05649) (119911 + 05556)
(119911 minus 05924) (119911 minus 09502) (119911 minus 01107))
)
119867PSO 3rd opt(119911)
1198673rd-fo-3(119911) = (093046 (119911 minus 01852) (119911 + 02354) (119911 + 0433)
(119911 minus 0742) (119911 + 03892) (119911 + 02492))
1198673rd-fo-6(119911) = (
(093038 (119911 minus 06395) (119911 + 04678) (119911 + 03939)
(119911 + 02476) (119911 + 0223) (119911 minus 0008033))
((119911 minus 09031) (119911 minus 02946) (119911 + 02512)
(119911 + 02397) (119911 + 0383) (119911 + 04194))
)
Table 6 Comparisons of RME values of the proposed half integrators based on optimized DIs with the existing models
120596
(120587 rad)
Proposed half integrators based on 2nd orderPSO optimized DI
Proposed half integrators based on 3rdorder PSO optimized DI
Gupta-Madhu-Jain halfintegrator
Krishna halfintegrator
1198672nd-fo-2(119911) 1198672nd-fo-4(119911) 1198672nd-fo-6(119911) 1198673rd-fo-3(119911) 1198673rd-fo-6(119911)
025 minus1579 minus2455 minus4325 minus1579 minus2926 minus3553 minus174205 minus1752 minus3874 minus5585 minus1752 minus3897 minus5362 minus2831075 minus2582 minus4156 minus8514 minus2581 minus4261 minus5365 minus355610 minus583 minus583 minus5544 minus6401 minus7475 minus4526 minus4205125 minus3125 minus5734 minus5348 minus3113 minus5067 minus3893 minus3713175 minus2894 minus8115 minus7532 minus2982 minus5825 minus3359 minus405720 minus3262 minus7583 minus5554 minus3262 minus7325 minus3488 minus4646225 minus4143 minus5713 minus4899 minus4115 minus5893 minus4369 minus5528275 minus3299 minus596 minus596 minus3633 minus4694 minus326 minus29830 minus2956 minus5603 minus4775 minus2827 minus4674 minus3374 minus243
orders and very lessmagnitude errors as compared to existingones Proposed FOIs based on the optimizedDIs present veryless RME values of the order of minus40 dB and show linear phasecurves in almost full band of Nyquist frequency
References
[1] L Debnath ldquoRecent applications of fractional calculus toscience and engineeringrdquo International Journal of Mathematicsand Mathematical Sciences vol 2003 no 54 pp 3413ndash34422003
[2] B T Krishna ldquoStudies on fractional order differentiators andintegrators a surveyrdquo Signal Processing vol 91 no 3 pp 386ndash426 2011
[3] N M F Ferreira F B Duarte M F M Lima M G Marcosand J A T Machado ldquoApplication of fractional calculus in thedynamical analysis and control of mechanical manipulatorsrdquoFractional Calculus and Applied Analysis vol 11 pp 91ndash1132008
[4] S Manabe ldquoThe noninteger integral and its application tocontrol systemsrdquo English Translation Journal Japan vol 6 pp83ndash87 1961
[5] J A T Machado and A M S Galhano ldquoA new methodfor approximating fractional derivatives application in non-linear controlrdquo in Proceedings of the 6th EUROMECHNonlinearDynamics Conference (ENOC rsquo08) pp 4ndash8 Saint PetersburgRussia July 2008
[6] Y Ferdi J P Herbeuval A Charef and B Boucheham ldquoR wavedetection using fractional digital differentiationrdquo ITBM-RBMvol 24 no 5-6 pp 273ndash280 2003
[7] J Mocak I Janiga M Rievaj and D Bustin ldquoThe use of frac-tional differentiation or integration for signal improvementrdquoMeasurement Science Review vol 7 pp 39ndash42 2007
[8] B Mathieu P Melchior A Oustaloup and C Ceyral ldquoFrac-tional differentiation for edge detectionrdquo Signal Processing vol83 no 11 pp 2421ndash2432 2003
[9] Y Pu W Wang J Zhou Y Wang and H Jia ldquoFractionaldifferential approach to detecting textural features of digital
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Journal of Optimization
image and its fractional differential filter implementationrdquoScience in China F vol 51 no 9 pp 1319ndash1339 2008
[10] S Das I Pan S Das andA Gupta ldquoImprovedmodel reductionand tuning of fractional-order PI120582D120583 controllers for analyticalrule extraction with genetic programmingrdquo ISA Transactionsvol 51 no 2 pp 237ndash261 2012
[11] S Das I Pan K Halder S Das and A Gupta ldquoImpact offractional order integral performance indices in LQR basedPID controller design via optimum selection of weightingmatricesrdquo in Proceedings of the International Conference onComputer Communication and Informatics (ICCCI rsquo12) pp 1ndash6Coimbatore India January 2012
[12] Y Chen and B M Vinagre ldquoA new IIR-type digital fractionalorder differentiatorrdquo Signal Processing vol 83 no 11 pp 2359ndash2365 2003
[13] Y Chen B M Vinagre and I Podlubny ldquoContinued frac-tion expansion approaches to discretizing fractional orderderivatives-an expository reviewrdquo Nonlinear Dynamics vol 38no 1ndash4 pp 155ndash170 2004
[14] Y Q Chen and K L Moore ldquoDiscretization schemes forfractional-order differentiators and integratorsrdquo IEEE Transac-tions on Circuits and Systems I vol 49 no 3 pp 363ndash367 2002
[15] M A Al-Alaoui ldquoNovel class of digital integrators and differ-entiatorsrdquo IEEE Transactions on Signal Processing p 1 2009
[16] M A Al-Alaoui ldquoAl-Alaoui operator and the new transfor-mation polynomials for discretization of analogue systemsrdquoElectrical Engineering vol 90 no 6 pp 455ndash467 2008
[17] M Gupta M Jain and N Jain ldquoA new fractional orderrecursive digital integrator using continued fraction expansionrdquoin Proceedings of the India International Conference on Powerelectronics (IICPE) pp 1ndash3 New Delhi India 2010
[18] M Gupta P Varshney and G S Visweswaran ldquoDigitalfractional-order differentiator and integrator models based onfirst-order and higher order operatorsrdquo International Journal ofCircuitTheory and Applications vol 39 no 5 pp 461ndash474 2011
[19] F Leulmi and Y Ferdi ldquoAn improvement of the rationalapproximation of the fractional operator 119904
120572rdquo in Proceedingsof the Saudi International Electronics Communications andPhotonics Conference (SIECP rsquo11) pp 1ndash6 Riyadh Saudi Arabia2011
[20] Y Ferdi ldquoComputation of fractional order derivative andintegral via power series expansion and signal modellingrdquoNonlinear Dynamics vol 46 no 1-2 pp 1ndash15 2006
[21] G S Visweswaran P Varshney and M Gupta ldquoNew approachto realize fractional power in z-domain at low frequencyrdquo IEEETransactions on Circuits and Systems II vol 58 no 3 pp 179ndash183 2011
[22] P Varshney M Gupta and G S Visweswaran ldquoSwitchedcapacitor realizations of fractional-order differentiators andintegrators based on an operator with improved performancerdquoRadioengineering vol 20 no 1 pp 340ndash348 2011
[23] B T Krishna and K V V Reddy ldquoDesign of fractional orderdigital differentiators and integrators using indirect discretiza-tionrdquo An International Journal for Theory and Applications vol11 pp 143ndash151 2008
[24] B T Krishna and K V V Reddy ldquoDesign of digital differentia-tors and integrators of order 12rdquo World Journal of Modellingand Simulation vol 4 pp 182ndash187 2008
[25] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization approachrdquoin Proceedings of the 2nd International Conference on Advances
in Recent Technologies in Communication and Computing (ART-Com rsquo10) pp 126ndash130 Kottayam India October 2010
[26] R Yadav and M Gupta ldquoDesign of fractional order differen-tiators and integrators using indirect discretization schemerdquoin Proceedings of the India International Conference on PowerElectronics (IICPE rsquo10) New Delhi India January 2011
[27] G E Carlson and C A Halijak ldquoSimulation of the fractionalderivative operator (1119904)
12 and the fractional integral operator(1119904)12rdquo Kansas State University Bulletin vol 45 pp 1ndash22 1961
[28] S K Steiglitz ldquoComputer-aided design of recursive digitalfiltersrdquo IEEE Transactions Audio and Electroacoustics vol AU-18 no 2 pp 123ndash129 1970
[29] A N Khovanskii The Application of Continued Fractions andtheir Generalizations to Problems in Approximation Theory PNoordhoff Ltd 1963 Translate by Peter Wynn
[30] S C Roy ldquoSome exact steady-state sinewave solutions ofnonuniform RC linesrdquo British Journal of Applied Physics vol 14pp 378ndash380 1963
[31] A Oustaloup F Levron B Mathieu and F M NanotldquoFrequency-band complex noninteger differentiator charac-terization and synthesisrdquo IEEE Transactions on Circuits andSystems I vol 47 no 1 pp 25ndash39 2000
[32] G Maione ldquoLaguerre approximation of fractional systemsrdquoElectronics Letters vol 38 no 20 pp 1234ndash1236 2002
[33] N Papamarkos and C Chamzas ldquoA new approach for thedesign of digital integratorsrdquo IEEE Transactions on Circuits andSystems I vol 43 no 9 pp 785ndash791 1996
[34] M Gupta M Jain and B Kumar ldquoRecursive wideband digitalintegrator and differentiatorrdquo International Journal of CircuitTheory and Applications vol 39 no 7 pp 775ndash782 2011
[35] N Jain M Gupta and M Jain ldquoLinear phase second orderrecursive digital integrators and differentiatorsrdquo Radioengineer-ing Journal vol 21 pp 712ndash717 2012
[36] C-C Hse W-Y Wang and C-Y Yu ldquoGenetic algorithm-derived digital integrators and their applications in discretiza-tion of continuous systemsrdquo in Proceedings of the CEC Congresson Evolutionary Computation pp 443ndash448 Honolulu HawaiiUSA 2002
[37] S Das B Majumder A Pakhira I Pan S Das and AGupta ldquoOptimizing continued fraction expansion based iirrealization of fractional order differ-integrators with geneticalgorithmrdquo in Proceedings of the International Conference onProcess Automation Control and Computing (PACC rsquo11) pp 1ndash6Coimbatore India July 2011
[38] G K Thakur A Gupta and D Upadhyay ldquoOptimizationalgorithmapproach for errorminimization of digital integratorsand differentiatorsrdquo in Proceedings of the International Con-ference on Advances in Recent Technologies in Communicationand Computing (ARTCom rsquo09) pp 118ndash122 Kottayam IndiaOctober 2009
[39] M A Al-Alaoui ldquoClass of digital integrators and differentia-torsrdquo IET Signal Processing vol 5 no 2 pp 251ndash260 2011
[40] D K Upadhyay and R K Singh ldquoRecursive wideband digitaldifferentiator and integratorrdquo Electronics Letters vol 47 no 11pp 647ndash648 2011
[41] D K Upadhyay ldquoClass of recursive wide band digital differen-tiators and integratorsrdquo Radioengineering Journal vol 21 no 3pp 904ndash910 2012
[42] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 December 1995
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Optimization 11
[43] Y-T Kao and E Zahara ldquoA hybrid genetic algorithm andparticle swarm optimization formultimodal functionsrdquoAppliedSoft Computing Journal vol 8 no 2 pp 849ndash857 2008
[44] Y Shi and R C Eberhart ldquoEmpirical study of particle swarmoptimizationrdquo in Proceedings of the IEEE international Con-ference Evolution Computer pp 1945ndash1950 Washington DCUSA 1999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of