Development of a High-sensitivity Accelerometer Board for Structural Health Monitoring

Hongki Jo,a* Jennifer A. Rice,b B.F. Spencer, Jr.,a and Tomonori Nagayamac

a Department of Civil and Environmental Engineering

University of Illinois, Urbana IL 61801, USA

b Department of Civil and Environmental Engineering Texas Tech University, Lubbock TX 79409, USA

c Department of Civil Engineering University of Tokyo, Tokyo 113-8656, Japan

ABSTRACT

State-of-the-art wireless smart sensor technology enables a dense array of sensors to be distributed through a structure to provide an abundance of structural information. However, the relatively low resolution of the MEMS sensors that are generally adopted for wireless smart sensors limits the network’s ability to measure low-level vibration often found in the ambient vibration response of building structures. To address this problem, development of a high-sensitivity acceleration board for the Imote2 platform using a low-noise accelerometer is presented. The performance of this new sensor board is validated through extensive laboratory testing. In addition, the use of the high-sensitivity accelerometer board as a reference sensor to improve the capability to capture structural behavior in the smart sensor network is discussed.

Keywords: high-sensitivity, wireless smart sensor, structural health monitoring (SHM), low-noise sensor

1. INTRODUCTION

The shift of structural health monitoring (SHM) research away from traditional wired methods toward the use of wireless smart sensor networks (WSSN) has been motivated by the many attractive features of a smart sensor. Recent state-of-the-art sensor technology enables wireless smart sensors having wireless communication, on-board computation, relatively low cost, and small size. These features enable the deployment of a dense array of sensors on structures, which can provide useful information and increase the potential of the SHM.

Several important factors determine the level of success that may be achieved by vibration-based SHM using smart sensors. First, a stable and reliable smart sensor network is required, which may be obtained through the advanced hardware and advanced networking software. Second, effective data processing techniques should be available to process the data using the on-board computation capabilities of a smart sensor. These goals may be achieved by the Imote2[1], which has been shown to be well suited for the application of a range of data aggregation and SHM algorithms (Rice, et al. 2010; Jang, et al. 2010; Cho, et al. 2010, Nagayama, et al. 2010). [2~5] Although the Imote2’s power consumption is higher than other smart sensors, such as the Mica2[6], which is specially focused on low power applications, careful power management utilizing a deep sleep mode[7] and energy harvesting efforts effectively elongates the life time of WSSN.[8]

Another fundamental factor for successful vibration-based SHM is that a high quality data should be obtained. If the measured data is contaminated with noise, the results from the SHM system will be unreliable, therefore undermining efforts to achieve successful SHM. In this paper, the development of a high-sensitivity sensor board (SHM-H board) for Imote2 platform is reported. This sensor board enables collection of high fidelity acceleration data, specifically focusing on low-level accelerations. The performance of the sensor board is compared with the commercially available ITS400C sensor board[9] and the SHM-A board developed by Rice and Spencer[7], and validated through extensive static and dynamic testing. Subsequently, the potential use of the SHM-H board as a reference to better utilize data taken by other lower-quality sensors in a network is explored.

Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2010, edited by Masayoshi Tomizuka, Chung-Bang Yun, Victor Giurgiutiu, Jerome P. Lynch, Proc. of SPIE

Vol. 7647, 764706 · © 2010 SPIE · CCC code: 0277-786X/10/$18 · doi: 10.1117/12.848905

Proc. of SPIE Vol. 7647 764706-1

2.1 Imot

The ImoXscale prup to 41SDRAMsupports nine diffapplicatiothe systedebugginsystem p

Figure

2.2 ITS4

Because Imote2. Ta three-asensor (T

The mainaxes digiBecause efficiencyapplicatio

i.

ii.

iii

iv

Even thosamplingcost is re

te2

ote2 (IPR2400rocessor, whic16MHz for c and 32MB a 250KB/s d

ferent voltageons. The Imotem for a specng through JTAower.

1. Imote: top

400 Basic Sen

the Imote2 haThe ITS400 baxis digital acTSL2561[12]), a

n sensor on thital output linthere is no ney of the senons:

Lack of theand corres70Hz cutof

. The samplcritical forsampling rboards.

i. The non-streported by

v. The resolutfor ambien

ough Nagayamg rate of the ITquired and the

2. W

0) is an advanch has a softwalculation intof flash memata rate, with

e domains andte2 is a modulcific applicati

TAG interface

(left), right (m

nsor Board

as no sensingasic sensor boccelerometer and a 4-ch 12-

he ITS400 bonear acceleroeed for a separnsor board. H

e flexibility inponding cutoff frequency),

ling rates are r SHM applicrate varied fro

tationary flucy Nagayama e

tion of the dignt vibration me

ma et. al.[10] cTS400 basic see limited reso

WIRELESS

nced smart seware selectabletensive tasks.mory. It also

an onboard 2d includes a lar stackable pion. The exteand USB por

middle), and b

g capabilities, oard develope(LIS3L02DQ-bit analog-to

ard is the ST meter havingrate ADC, theHowever the

n selecting theoff frequencie

which is too h

not stable, wcation.[14] Tes

om 537Hz to 6

ctuation in thet. al.[14]

gital acceleromeasurements.

could effectivensor board wlution remain

SMART SE

ensor platforme CPU speed t The platformintegrates an

2.4GHz antenbattery chargplatform and

ension boards rt programmin

asic sensor bo

a separate send by Intel, and

Q, ST Micro[1

-digital conve

Micro LIS3Lg a built-in 12e LIS3L02DQ

LIS3L02DQ

e sampling rats. The loweshigh for many

which can signsts conducted605Hz for the

he sampling r

meter with the

vely resolve twith the use ofs a critical iss

ENSOR HAR

m. It is built that ranges from integrates n 802.15.4 Tnna. The Poweging option, wcan be extendinclude the

ng, and the ba

oard stacked o

nsor board is d commercial10]), a temp/herter (ADC; M

L02DQ accele2 bit ADC w is a low-pow

acceleromet

te. The ITS400t sampling ra

y SHM applic

nificantly hind by Nagayame expected sam

rate on the b

e built-in ADC

the synchronif a resamplingue.

RDWARE

around Intelom 13MHz fo256KB of SI CC2420 raer Manageme

which is usefuded with diverInterface boa

attery board (I

on Imote2 and

required for tly available fr

humidity sensoMaxim 1363[13

erometer. The which is follower sensor, conter has some

0 board has oate is 280Hz ations.

nder synchronma and Spencmpling rate of

asic sensor b

C is just 0.98m

ization probleg algorithm, ad

’s low-poweror low power oRAM, 32MB

adio transceivent IC (PMICful for long-terse boards to card (IIB2400)IBB2400CA)

d battery board

to collect datarom MEMSICor (SHT15[11]

]).

LIS3L02DQ owed by digitntributing to th limitations

nly four samp(with a corre

nized sensing cer[15] showedf 560Hz with

board up to 0

mg, which is t

em due to thedditional comp

r PXA271 operation,

B external ver, which ) supplies erm SHM customize for code to supply

d (right).

a with the C employs ]), a light

is a three tal filters. he power-for SHM

pling rates esponding

that so is d that the 14 sensor

0.1% was

too coarse

e unstable putational

Proc. of SPIE Vol. 7647 764706-2

2.3 UIU

Rice andthe ITS4user-seleprogrammrelated prI/O. The

i.

ii.

iii.

iv.

PG

UC SHM-A se

d Spencer[7] de00. Structuralctable samplinmable digital roblems. The key features o

AcceleromST M0.66and

QuickfilterTheprogfiltealiarequfilte

OP Amp Eacextethe attenimp

ResolutionThemax0.430.670.43

GA 3

ensor board

eveloped an al Health Moning rates throufilters. CombiSHM-A board

of the SHM-A

Figure 2

meter Micro LIS3446V/g sensitivi50μg/√Hz no

r ADC e key compogrammable 4-ering, and desing filter (Auirements. Thers.

Figure 3. F

ch channel of ernal low-passinput voltagenuation. An

pedance and el

n & Noise of Se effective nuximum resolu3mg. Consider7mg for the z3mg for x- and

AA Filter

3rd Order Besse fc = 500kHz

accelerometer itoring Accele

ugh the use of ining an analod has 3 axes a

A sensor board

2. SHM-A boa

4ALH[16] anality and 0.68moise density.

onent of the -channel ADCcimation and

AA filter), cuis functionalit

Functional Blo

the LIS344ALs filters; the Qe levels to the

operational liminate the si

SHM-A[7] umber of bits tion that the Qring the avera

z-axis over a 2d y-axes and 0

el

1

board for theerometer (SHMf the Quickfiltog acceleromeanalog MEMSd are summariz

ard: Top (le

log 3 axes accmA current con

SHM-A boC and signal cod can be indivutoff frequencty is accompli

ock Diagram o

LH acceleromQF4A51 has ae Quickfilter A

amplifier (Tignal attenuati

(ENOB) of Quickfilter ADage RMS noise20Hz bandwid0.67mg for the

16-bitADC

e Imote2 platfM-A) board per QF4A512, eter with the OS accelerometezed below.

ft) and bottom

celerometer. Tnsumption at

oard is the Qonditional . Evidually progcy, ADC samished with fou

of Quickfilter

meter has an ina 10KΩ input ADC. This m

TI OPA4344[1

ion.

the QF4A51 DC can achiee of the SHM-dth, the actuae z-axis (0~20

Cascaded InComb (CIC

Cascaded InHalfband

Filter/Dec

form to addreprovides flexib

which is a 4-OF4A512 has er and interfac

m (right).

This low-cost 3.3VDC pow

Quickfilter QEach channel pgrammed for mpling frequeur separate hig

r QF4A512.

nternal 110KΩimpedance fo

mismatch resul18]) is emplo

is 13.2bit.[17]

eve for a ±2g -A, 0.29mg foal resolution o0Hz).

ntegrator-C) and

ntegrator-(CIH)

imator

ess the shortcoble and highly-channel ADC

resolved samces with Imote

accelerometerwer, ±2g sensi

QF4A512.[17] performs overthe gain (PG

ency, and unigh-precision d

Ω resistor as por each channlts in significoyed to prov

] With the ENsensing range

or the x- and yof the SHM-A

Programma

FIR Filte

omings of y accurate

C and with mpling rate e2 via SPI

r provides ing range,

It is a rsampling, GA), anti-ique filter digital FIR

part of an el to limit ant signal vide high

NOB, the e is about

y-axes and A board is

able

r

Proc. of SPIE Vol. 7647 764706-3

3. HIGH-SENSITIVITY SENSOR BOARD 3.1 Design criteria

Strong structural excitations, such as earthquakes and hurricanes, can result in high levels of structural response which are readily captured by sensors with limited measurement resolution. In particular, because the signal-to-noise ratio (SNR) of the structural responses is high, the measurement noise becomes negligible. However, most structural responses that can be measured during routine monitoring are low-level ambient vibration responses. Ambient vibration is generated due to a variety of random excitation sources such as nearby traffic, normal wind-loading, machinery inside the structures, etc. Ambient vibration data can be useful for vibration-based SHM. However, usually the vibration level is too small to capture with commercially available sensor board for wireless sensors. The nominal resolution of the SHM-A board is 0.43~0.67mg, which is about twice better than the ITS400C basic sensor board ADC resolution; however it is not able to measure the ambient vibration levels of less than 1mg that are produced by most of civil structure.

Efforts to improve the data quality from the sensor board require consideration of several factors[7] First, the noise floor of the sensor and other electrical components should be sufficiently low. Second, the sensitivity of the sensor, which is the relationship between the physical phenomena and output of the sensor, should be sufficiently high. The final factor is that the resolution at which the analog signal is digitized by the ADC should be fine enough.

A new high-sensitivity sensor board has been developed for measuring low-level ambient vibrations of structures. The SHM-H board is the extension version of the SHM-A board (general purpose acceleration board). The SHM-H board also employs the Quickfilter ADC; however, the z-axis of the SHM-A board, which has the highest noise among the 3-axes, is replaced with a low-noise and high-sensitivity sensor. In addition, by reducing the span of the ADC, the resolution of the ADC can be improved. Figure 4 and Figure 5 show the block diagram and sensor board design, respectively.

Figure 4. Block diagram of SHM-H board.

Figure 5. Top view (left), bottom view (middle) of SHM-H board

Proc. of SPIE Vol. 7647 764706-4

3.2 Low-noise accelerometer (SD1221)

For the high sensitivity-acceleration board, the Silicon Designs SD1221L-002[19] low-noise accelerometer is used. This accelerometer is a micro-machined capacitive-type sensor with ±2g and DC to 400Hz sensing range. In particular, the 2g version has a noise density of 5µg/√Hz and a sensitivity of 2000mV/g with the differential analog outputs. The SD1221 contains a temperature dependent current source that is useful for measuring the internal temperature of the accelerometer so that any previously characterized bias and scale factor temperature dependence can be corrected. The main features of the SD1221 are given in Table 1.

Table 1. SD1221L-002 accelerometer specifications. Parameter Value

Input Range ±2g Frequency Response 0 ~ 400 Hz Sensitivity (Differential) 2000 mV/g Output Noise (Differential, RMS) 5 µg/√Hz (10 µV/√Hz)_ Bias Temperature Shift 0.4 mg/°C (max 1.2 mg/°C) Scale Factor Temperature Shift -1 ~ +1 mg/°C Output Impedance 90 Ohms Operating Voltage Typ. 5 Volts (4.75 ~ 5.25 Volts) Operating Current Typ. 8 mA (max 10 mA)

3.3 Low-noise power supply

A stable and clean power supply is critical for a low-noise sensor. Because the output signal of the SD1221 is ratiometric to the 5V power supply, if the power supply is noisy, then the output signal also is noisy. The Power Management Integrate Circuit (PMIC) of the Imote2 provides regulated 1.8V and 3.0V for powering sensor boards and also provides a 5V power source. However, because the 5V is generated by the voltage booster used in the USB host controller on the Imote2, it does not provide a power supply that is adequate for the low-noise sensor. In addition, the SD1221 requires a limited power supply range of 4.75 ~ 5.25V. To achieve a clean power supply within the required range, the 5V power supply from the Imote2 is regulated utilizing the MAX 8878[20] low-noise and low-dropout linear regulator. The output noise of the regulator is only 30 μVrms over 10Hz to 100KHz (≈ 0.1µV/√Hz), and the dropout is just 55mV at 50mA output.

3.4 Sensing range reduction

The SNR of the Quickfilter ADC specified in the datasheet that results from noise within the device is given as 81dB, which corresponds to 13.2 effective number of bits (ENOB). Equation 1 gives the relation between ENOB (in bits) and SNR (in dB).

(1) 1.766.02

If the full scale of the SD1221L-002, ±2g, is used, the resolution that the ADC can achieve is just 4000 mg /2 . 1 0.43mg with the ENOB of the ADC. For the ambient vibration of 1mg level, the resolution of 0.43mg is not sufficient. By limiting the measurement range of the sensor to ±0.2g for horizontal acceleration or +0.8g ~ +1.2g for vertical acceleration, a maximum resolution of 400 mg / 2 . 1 0.043mg is achieved which is sufficient to capture low-level acceleration in the range of 1~2mg.

To reduce the sensing range, the output signal needs to be amplified and shifted. Using the OP-Amp circuit shown in Figure 6, the differential output signals from the accelerometer can be easily controlled.

Proc. of SPIE Vol. 7647 764706-5

Figure 6. OP-Amp circuit for the amplification and shift of the signal.

3.5 Noise characteristic of OP-Amp The raw output signals from the SD1221 are amplified and shifted by the OP-Amp circuit shown in Figure 6 to achieve improved ADC resolution. The noise from the OP-Amp also should be limited. To estimate the OP-Amp noise effects on the signal, the interaction between OP-Amp voltage noise ( ), OP-Amp current noise ( ) and resistor noise ( ) must be understood. The total input referred noise ( ) of an OP-Amp is given by Equation 2.

/ (2)

where, is the total equivalent source resistance at the two inputs, and 4 0.13 in nV/√Hz at 25 1.38 10 J/K Boltzmann′s constant

For the TI OPA4344,[18] the voltage noise is 30 nV/√Hz, and the current noise is 0.5 fA/√Hz over 10KHz. The resulting total noise is 31.4 nV/√Hz with 5.2KΩ, which is the equivalent resistance of the SHM-H board’s OP-Amp circuit. In comparison with the output noise of SD1221 (10 µV/√Hz) shown in Table 2, the OP-Amp noise 31.4 nV/√Hz is negligible.

3.6 Low-noise printed circuit board (PCB) design The separation of the analog and digital sections of the PCB keeps the noisy digital signals away from the low level analog signals. The SHM-H board has a mostly solid ground plane on the bottom side, which is split between the digital and the analog ground plane. The digital signal lines are on the digital plane and the analog signal lines are over the analog ground plane. It is important that the analog ground and the digital ground be connected together at the ADC. This approach allows a quick return for the ground currents as the analog and digital portions of the device communicate. The bridge to connect two ground planes is just below the Quickfilter ADC.

4. TESTING AND VALIDATION

4.1 Static noise characteristic test To estimate the actual noise floor of the SHM-H board and compare with the performance of the other MEMS accelerometers, static noise characteristic tests were conducted. The SHM-H board resting on an aluminum plate was placed on the desk in the basement of the Newmark Civil Engineering Laboratory. In total 20,000 data points were measured at 280Hz (70Hz cutoff frequency) using the RemoteSensing application of ISHMP Services Toolsuite.[21] Data from the SHM-H board was compared with the data from the SHM-A board and the ITS400C basic sensor board under the same conditions.

The test results given in Figure 7 show the clear difference between the three sensor boards in the time and frequency domain. The RMS noise over a given bandwidth is obtained by taking a square root of the area under

Proc. of SPIE Vol. 7647 764706-6

the PSD curve within the bandwidth. For the ITS400C sensor board, signal quantization was observed (see the left of Figure 7), resulting from the low resolution of the digital accelerometer’s built-in ADC. For the ±2g sensing range, the 12-bit ADC provides only 4000 mg / 2 1 0.98mg resolution. Though a test showed the RMS noise of the ITS400C sensor board over 20Hz bandwidth was 0.3mg (corresponding noise density: 67.1µg/√Hz), the actual resolution of the sensor board is determined by the ADC resolution of 0.98mg. Also, the stop band attenuation is not clear, as shown in the power spectral density (PSD) of the ITS400C sensor board (see the right of Figure 7). For the SHM-A board, there is no observable quantization and the stop band attenuation is clear at the specified 70Hz cutoff frequency. Rice and Spencer have reported the RMS noise of 0.29mg for the x- and y-axes and 0.67mg for the z-axis over a 20Hz bandwidth through hundreds tests[7], which is about the twice of the ITS400C’s resolution. The SHM-H board shows quite promising results, even though the results come from a proto-type version of SHM-H board. The RMS noise over a 20Hz bandwidth is just 0.06 mg, which is about 10 ~ 20% of SHM-A board’s noise level.

Figure 7. An example of static test result: time histories (left) and power spectral densities (right).

Table 2. Resolution comparison over 20Hz bandwidth of sensor boards for Imote2 platform.

RMS noise Noise density ADC resolution Final resolution ITS400C 0.3mg 67.1 µg/√Hz 0.98mg 0.98mg

SHM-A board 0.29mg(x/y axis) ~ 0.67mg(z axis) [7] 62.6 µg/√Hz 0.43mg 0.43mg(x/y axis) ~

0.67mg(z axis) SHM-H board 0.06 mg 13.4 µg/√Hz 0.043mg 0.06mg

4.2 Dynamic shaking table test The SHM-H board was tested on a vertical shaker to check that the sensor had predictable response and assess how well the response matches with the response from a conventional accelerometer when subjected to dynamic motion. For this test, an electrodynamic shaker (LDS V408)[22] and a piezo-electric type PCB393C[23] accelerometer as a reference sensor were used. The PCB393C has 0.1mg resolution over 1~10,000 Hz broad band, which corresponds to 1 µg/√Hz noise density roughly.

Figure 8. Electrodynamic shaker (left) and PCB393C accelerometer (right).

0 10 20 30 40 50 60 70-3

-2

-1

0

1

2

3

time(s)

acce

lera

tion(

mg)

0 20 40 60 80 100 120 140

10-6

10-4

10-2

100

frequency(hz)

mg2 /H

z

Basic sensor boardSHM-A board (ST-Micro)SHM-H board (SD1221)

Basic sensor boardSHM-A board (ST-Micro)SHM-H board (SD1221)

Proc. of SPIE Vol. 7647 764706-7

A 50Hz band-limited white noise excitation was used for the comparison of the SHM-H board and the PCB393C, and the results given in Figure 9 show excellent agreement between the wired sensor and the SHM-H board in the time and frequency domain.

(a)

(b)

(c)

Figure 9. Shaker test results in time- and frequency domain for 50Hz band-limited white noise excitation: (a) Time histories (left) and zoomed (right), (b) power spectral densities (left) and coherence (right), and (c) transfer function (left) and phase (right).

It should be noted that the shaker has a limited ability to generate energy in lower frequency range and the wired sensor does not have DC measurement capability, so that slight disagreements in the lower frequency range were observed.

5. TEMPERATURE CORRECTION A capacitive type sensor is susceptible to mean value drift of the signal due to the temperature change inside the sensor. The SD1221 accelerometer also experiences this drift effect; according to the datasheet, the bias temperature drift is about 0.4 mg/°C. Usually there is no need to consider this issue when measuring relatively short periods of data, because the temperature change inside the sensor is not that significant. However, long data records will exhibit drift in the constant acceleration voltage. Figure 10 shows the time history of a constant acceleration signal measured by the SHM-H board for 2000 seconds; as is seen, the signal drifts from -1mg ~ +2.5mg.

0 0.5 1 1.5 2 2.5 3 3.5 4-300

-200

-100

0

100

200

300

mg

Synchronized Time Histories after Synch

PCB393CSHM-H

2.5 2.6 2.7 2.8 2.9 3-200

-100

0

100

200

Time (sec)

mg

Synchronized Time Histories - zoomed

PCB393CSHM-H

0 10 20 30 40 50

-40

-20

0

20

40

60

dB

PSD

PCB393CSHM-H

0 10 20 30 40 500

0.2

0.4

0.6

0.8

1

f (Hz)

Coherence

0 10 20 30 40 50-25

-20

-15

-10

-5

0

5Transfer Function

dB

0 10 20 30 40 50-60

-40

-20

0

20

40

60Phase

freq (Hz)

degr

ees

Proc. of SPIE Vol. 7647 764706-8

Figure 10. Mean-value drifted acceleration signal (2000 seconds measurement).

Although this drift usually does not affect the frequency content of the original signal, it is essential to consider in the case that the magnitude of the acceleration is important.

The SD1221 provides a temperature-dependent current source output. This temperature-dependent current is useful for the estimation of the temperature inside the sensor so that the associated drift can be corrected. The nominal output current at 25°C is ≈ 500µA and the nominal sensitivity is 1.5µA/°C. The current signal, however, should be converted to voltage signal that the Quickfilter ADC can read. The 4th channel of the ADC in the SHM-H board is used to measure this temperature dependent voltage signal. Figure 11 shows the OP-Amp circuit for converting the current signal to the voltage signal. For the temperature range of -15°C ~ +85°C, the default setting of the circuit (R1 = 20K, R2 = 3.6K, and Vr = 1.8V) provides the voltage signal change of 3V ~ 0V.

Figure 11. OP-Amp circuit for the signal converting (current voltage).

Figure 12(b) shows the raw data of the temperature using the OP-Amp circuit shown in Figure 11; as is observed, the temperature signal has quite similar shape with the drifted acceleration signal shown in Figure 12(a); the linear relation is clearly apparent in Figure 12(c). Using the scaled temperature signal shown in Figure 12(d), the drifted acceleration signal can be corrected. The temperature corrected acceleration signal and its power spectrum are shown in Figure 13. It should be noted that the use of the scaled-smoothed temperature signal, which is just temperature change trend, gives better correction than the use of just scaled raw temperature signal. Because the temperature signal itself can have some noise, it can add noise on the corrected acceleration signal. Figure 13(b) shows that the temperature correction using the scaled temperature data itself increases the noise level little in the low frequency range (see dotted box in the PSD) while the scaled-smoothed temperature trend doesn’t affect on the PSD.

0 200 400 600 800 1000 1200 1400 1600 1800 2000-2

-1

0

1

2

3

time(s)

Acc

eler

atio

n (m

g)

(a) (b)

Proc. of SPIE Vol. 7647 764706-9

Figure 12. Relation between mean-value drift in acceleration signal and inside temperature: (a) raw acceleration

data from ADC, (b) raw temperature data from ADC, (c) linear relation between the acceleration & the internal temperature, (d) scaled- and scaled-smoothed temperature signal.

Figure 13. Time history (left) and power spectral density (right) of temperature corrected acceleration

6. APPLICATIONS OF THE SHM-H BOARD The objective of the SHM-H board is to effectively measure a low-level ambient vibration. The ultimate noise density of the SHM-H board is approximately 13.4µg/√Hz (RMS noise over 20Hz bandwidth: 0.06mg), which is small enough for measuring the ambient response of a structure (< 1mg). A test for the application to the low-level ambient vibration was performed with the truss structure in the Smart Structure Technology Laboratory (SSTL) in Newmark Civil Engineering Laboratory. The ambient excitations generated by nearby diverse experimental facilities and people walking are transferred to the truss through the supports at the both ends of the structure. Figure 14 shows the truss structure and the sensor locations for the testing.

Figure 14. Steel truss structure in SSTL (left) and sensor locations for testing (right). The test results given in Figure 15 shows that the ambient response level of the truss is about 1~2mg (left), the SHM-A board, which is a general purpose acceleration board, did not catch peaks around 10Hz, 33Hz, 40Hz, and 63Hz very well, while the SHM-H board and the PCB393C clearly found the peaks (right).

0 500 1000 1500 2000

-1

0

1

2

3

time(s)

Acc

eler

atio

n (m

g)

Acceleration-originalTemp. corrected Acc (using raw temp.)Temp. corrected Acc (using temp. trend)

0 10 20 30 40 50

10-6

10-4

10-2

100

frequency(hz)

mg2 /H

z

Acceleration-originalTemp. corrected Acc (using raw temp.)Temp. corrected Acc (using temp. trend)

(c) (d)

Proc. of SPIE Vol. 7647 764706-10

Figure 15. Ambient vibration on truss structure in SSTL: time histories (left), power spectral densities (right).

Another possible application of the high-sensitivity sensor board is that it can be used as a reference sensor in a sensor network. The combined use of a conventional wired sensor and smart sensors can improve the data quality of whole sensor network[24]. If the signal measured by a particular sensor is composed of the signal and sensor noise, and the noise and the signal are not correlated, the cross correlation (or cross spectrum) of the signal from a high-sensitivity sensor and the noisy signal from smart sensor node is expected to quickly approach the cross correlation (or cross spectrum) of the signals without observation noise:

from wireless smart sensors, high noise from wired conventional sensors, low noise where, measurement, signal, and noise

As indicated previously, the SHM-A board was not able to capture some of the peaks which appeared in the PSD of the SHM-H sensor board data. However, the cross power spectrum of the SHM-A board data and the SHM-H board can eliminate part of the noise in the measurements of the SHM-A board data so that the peaks are captured in the same way as they are captured in the PSD of the SHM-H sensor output (see Figure 16).

Figure 16. Cross power spectrum of SHM-A data and SHM-H data.

7. FUTURE WORK

The next plan in the development of high-sensitivity sensor boards suitable for SHM is to use another low-noise MEMS accelerometer, SF1500SA (Colibrys).[25] The noise density of the sensor is just 0.3 µg/√Hz. The relatively high cost of the sensor (~$480) is a potential drawback for its use with a wireless smart sensor, however it may be an attractive option as a less densely deployed reference sensor. Further evaluation of the SHM-H sensor board in a full-scale test setting is planned for the future.

8. CONCLUSIONS

In this paper, the development of a high-sensitivity sensor board for Imote2 platform was presented. Diverse considerations for the low-noise sensor board were addressed, and the performance was tested and validated through static and the dynamic testing. An experiment using a truss structure demonstrated that the SHM-H

0 10 20 30 40 50 60 70-4

-2

0

2

4Time history

acce

lera

tion(

mg)

time(sec)

SHM-A(21)SHM-A(25)SHM-H(19,SD1221)

0 20 40 60 80 100 120 140

10-5

100

PSD

frequency(hz)

mg2 /H

z

SHM-A(21)SHM-A(25)SHM-H(19,SD1221)PCB393-C(19,rear)

0 10 20 30 40 50 60 70 80

10-6

10-4

10-2

CPSD & PSD

frequency(hz)

mg2 /H

z

CPSD[SHM-A(21),SHM-H(19)]CPSD[SHM-A(25),SHM-H(19)]PSD[SHM-H(19)]

Proc. of SPIE Vol. 7647 764706-11

board could be effectively used for measuring the low-level ambient vibration of the structure. The potential for its use as a reference sensor to improve overall performance in a sensor network was also discussed.

9. ACKNOWLEDGEMENT

This study is supported in part by the National Science Foundation Grants CMS 06-00433 (Dr. S.C. Liu, program manager). This support is gratefully acknowledged.

10. REFERENCES

[1] MEMSIC, Inc, “IPR2400, Imote2 Wireless Sensor Node”, Andover, MA (2010) [2] Rice, J.A., Mechitov, K., Sim, S.H., Nagayama, T., Jang, S. A., Kim, R., Spencer, Jr., B. F., Agha, Gul, and

Fujino, Y., “Flexible Smart Sensor Framework for Autonomous Structural Health Monitoring,” Smart Structures and Systems (accepted) (2010).

[3] Jang, S.A., Jo, H., Cho, S., Mechitov, K., Rice, J.A., Sim, S.H., Jung, H.J., Yun, C.B., Spencer, Jr., B.F., and Agha, Gul, (2010) “Structural Health Monitoring of a Cable-stayed Bridge using Smart Sensor Technology: Deployment and Evaluation,” Smart Structures and Systems (accepted) (2010).

[4] Cho, S., Jo., H., Jang, S. A., Park, J., Jung, H. J., Yun, C. B., Spencer, Jr., B. F., and Seo, J. (2010). “Structural Health Monitoring of a Cable-stayed Bridge Using Smart Sensor Technology: Data Analyses,” Smart Structures and Systems (accepted) (2010).

[5] Nagayama, T., Moinzadeh, P., Mechitov, K., Ushita, M., Makihata, N., Ieiri, M., Agha, Gul, Spencer, Jr., B. F., Fujino, Y. and Seo, J. (2010). “Reliable Multi-hop Communication for Structural Health Monitoring,” Smart Structures and Systems (accepted) (2010).

[6] MEMSIC, Inc, “MICA2, Wireless Measurement System”, Andover, MA (2010) [7] Jennifer A. Rice and Billie F. Spencer, Jr., “Flexible Smart Sensor Framework for Autonomous Full-scale

Structural Health Monitoring,” NSEL Report, Series 018, University of Illinois at Urbana-Champaign, http://hdl.handle.net/2142/13635 (2009)

[8] T.I. Miller and B.F. Spencer Jr., “Solar Energy Harvesting and Software Enhancements for Autonomous Wireless Smart Sensor Networks,” NSEL Report, Series, University of Illinois at Urbana-Champaign, (2010)

[9] MEMSIC, Inc, “ITS400, Imote2 Basic Sensor Board”, Andover, MA (2010) [10] ST Microelectronics, “LIS3L02DQ MEMS Inertial Sensor,” Geneva, Switzerland (2005) [11] Sensirion, “SHT15 Temperature/Humidity sensor,” Switzerland (2005) [12] TAOS, “TSL2651 Light Sensor,” Plano, TX (2009) [13] Maxim, “Max1363 ADC,” San Jose, CA (2008) [14] T. Nagayama, J. A. Rice, and B. F. Spencer, Jr, “Efficacy of Intel’s Imote2 wireless sensor platform for

structural health monitoring application,” Proc. Asia-Pacific Workshop on Structural Health Monitoring, Yokohama, Japan (2006)

[15] Tomonori Nagayama and Billie F. Spencer, Jr., “Structural Health Monitoring using Smart Sensors,” NSEL Report, Series 001, University of Illinois at Urbana-Champaign, http://hdl.handle.net/2142/3521 (2007)

[16] ST Microelectronics, “LIS344ALH MEMS Inertial Sensor,” Geneva, Switzerland (2008) [17] Quickfilter Technologies, Inc., “QF4A512 4-channel programmable signal conditioner,” Allen, TX (2007) [18] Texas Instruments, Inc., “OP-amp4344”, Dallas, TX [19] Silicon Designs, Inc., “Model 1221 Low Noise Analog Accelerometer,” Issaquah, WA (2007) [20] Maxim, “Max8878 Low-dropout, low-noise Regulator,” San Jose, CA (2008) [21] University of Illinois, “Illinois Structural Health Monitoring Project (ISHMP),” Urbana, IL (2010),

http://shm.cs.uiuc.edu/ [22] LDS Test and Measurement. “Model LDS 408,” Herts, England (2007) [23] PCB Piezotronics Inc. “Model PCB 393C,” Depew, NY (2007) [24] Nagayama, T., Ushita, M., Fujino, Y., Ieiri, M., and Makihata, N. (2010). “The combined use of low-cost

smart sensors and high accuracy sensors to apprehend structural dynamic behavior,” Proc. SPIE Smart Structures/NDE (2010).

[25] Colibrys, Lnc., “SF1500S.A.,” Neuchatel, Switzerland (2009)

Proc. of SPIE Vol. 7647 764706-12