Journal Search Engine
Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 1598-4095(Print)
ISSN : 2287-7401(Online)
Journal of The korean Association For Spatial Structures Vol.18 No.4 pp.49-59
DOI : https://doi.org/10.9712/KASS.2018.18.4.49

Comparison Between Performance of a Wireless MEMS Sensor and an ICP Sensor in Shaking Table Tests

S. T. Mapungwana*, Young-Seok Jung**, Jong-Ho Lee***, Sung-Won Yoon****
*Dept. of Architecture, Seoul National University of Science and Technology
**I'ST Co., Ltd.
***Dept. of Architecture, Seoul National University of Science and Technology
****School of Architecture, Seoul National University of Science and Technology

Tel: 02-970-6587 Fax: 02-979-6563 E-mail: swyoon@seoultech.ac.kr
August 1, 2018 August 27, 2018 August 29, 2018

Abstract


Wireless sensors are more favorable in measuring structural response compared to conventional sensors. This is because they are easier to use with no issues with cables and are considerably cheaper. There are several applications that can be used in recording and analyzing data from MEMS sensor installed on an iPhone. The Vibration App is one of the applications used and there has not been adequate research conducted in analyzing the performance of this App. This paper analyzed the performance of the Vibration App by comparing it with the performance of an ICP sensor. Results show that natural frequency results are more accurate (error less than 5%) in comparison to the amplitude results. This means that built- in MEMS sensor in smartphones are good at estimating natural frequency of structures. In addition, it was seen that the results became more accurate at higher frequencies (5.0Hz and 10.0Hz).



진동대를 이용한 무선 MEMS 센서와 ICP 가속도계의 성능 비교

마 푼과나 시부시시웨*, 정 영 석**, 이 종 호***, 윤 성 원****
*학생회원, 서울과학기술대학교 건축과, 석사과정
**정회원, ㈜아이스트 대리
***학생회원, 서울과학기술대학교 건축과, 박사과정
****교신저자, 정회원, 서울과학기술대학교 건축학부 교수, 공학박사

초록


    1. Introduction

    Development of tall buildings has been increasing all over the world. Advances in height of buildings have led to an increase in the flexibility of these structures, which makes them more prone to horizontal loads1). Structural vibration of buildings due to horizontal loads such as strong winds continues to be an issue.

    Natural frequency of the structure is important in estimating the response of the structure and prediction of natural frequency of the structure is important in estimating the response acceleration of the structure2).

    MEMS (Micro Electro Mechanical Systems) sensors have been widely used in estimating the response acceleration of the structures. Conventional sensors have high installation requirements, maintenance costs, and issues with cables3). Unlike conventional sensors, MEMS sensors are cheaper and manufactured in smaller sizes. The performance and applicability of the commercial MEMS in building monitoring have been verified through vibration measurements4).

    The performance of MEMS sensor built in smartphones has been validated by a series of shaking table tests. An application called i-jshin for recording and analyzing the vibrational data was used for the test5).

    Comparison of the performance of MEMS sensor using shaking table tests in analyzing the performance of the sensor in terms of amplitude has been done. A piezoelectric sensor was used as the reference sensor and a Seismometer Application was used for the vibration measurements3).

    Analyzing of the performance of the MEMS sensor in terms of both natural frequency and amplitude using shaking table tests has been done. A servo-velocity sensor was used as the reference sensor and an application called Motion Graphs was used for vibration measurements6).

    It was seen that the observable acceleration level of smart devices is more than 5gal for frequencies in the range of 0.1 to 10Hz in shaking table tests6).

    The comparison of the performance of the Vibration App to I-jshin App has been conducted2). However, there has not been comparisons between the performance of a MEMS sensor using the Vibration App to the performance of a reference sensor.

    Comparing the performance of the Vibration App to a reference sensor is significant because reference sensors tend to be more accurate hence, the difference in accuracy can be easily observed. This study hopes to investigate the accuracy of using the Vibration App for measuring vibration response, in comparison to an ICP sensor.

    2. Vibration application

    Vibration is a spectrum analyzer of vibration that uses built in accelerometers and gyroscope inside an iPhone, iPod Touch, and iPad7). It acquires and displays time series data, and produces frequency spectra by performing an FFT on each channel7). As shown in <Fig. 1>, real-time measurements can be performed in three axes directions of x, y, z (two-direction horizontal acceleration and one-direction vertical acceleration).

    The 3-axis MEMS accelerometer (BMA280, Bosch Sensortech) can measure within a range of ± 2g8). In addition, the Vibration App can measure the sampling rate in the range of 0~ 100Hz. The dynamic characteristics of the sensor can be observed on the screen of the measuring device as shown in <Fig. 2> and <Fig. 3>.

    2.1 Time series view

    This view shows the time traces for all activated channels. The waveform due to vibration is shown in <Fig. 2 (a)>. Measurements are recorded for x, y and z-axis. The maximum, minimum, and rms values measured for each axis are calculated and displayed. The screen also displays the horizontal and vertical scale. The value of the horizontal scale is obtained from the sample rate and sample length set in the Data Acquisition Settings, and the value of the vertical scale is determined by the display setting of the time series waveform. In addition, the touch cursor helps in reading precise measurement values.

    <Fig. 2 (b)> shows a detailed waveform. Measurements can be initiated by pressing the sample button. The measured time series waveform and frequency waveform measurement data can be extracted as excel files for further analysis7).

    2.2 Frequency view

    This view displays the frequency content for each channel by carrying out a Fast Fourier Transform to the data from each channel. Similar to the time series view, the touch cursor can be used to precisely read measurements. The natural frequency can be analyzed in several ways according to the unit and function of the accelerometer as shown in <Fig. 3>7).

    2.3 Data acquisition settings

    <Fig. 4> shows the data acquisition screen. Sampling rate can be set and it should be at least twice the highest frequency to be measured7). In this experiment, a sampling rate was set to 100Hz.

    Sample delay is the number of seconds after pressing the sample button before data acquisition starts. This is helpful in positioning of the measuring devices7).

    Sample length enables the duration of acquisition to be set7). Sample length was set to 40.96 seconds for this experiment.

    3. Performance examination using shaking-table tests

    To investigate the performance of the Vibration Application in obtaining data from MEMS sensor, a series of shaking-table tests with varying frequencies and amplitudes were conducted.

    The equipment used for the shaking table test was a vibration shaker that sets the frequency and the acceleration level, a power amplifier that amplifies the input excitation, and a dynamic signal analyzer that converts the acceleration value of the accelerometer to an electric signal. The list of equipment used and their properties are given in <Table 1>. <Fig. 5> shows the equipment used.

    3.1 Specimen used

    ICP sensor and MEMS sensors were used in this experiment. ICP sensor was used as the reference sensor to compare results with the MEMS sensor. An iPad mini 4 was used for the Vibration Application experiment. The iPad mini 4 has a MEMS sensor and the Vibration Application installed on it. <Table 2> shows the characteristics of the two sensors used. The sufficient sampling frequency for MEMS sensor is 100Hz.

    3.2 Experimental setup

    <Fig. 6> shows layout of the specimen on the shaking-table test. Several specimen were used for this test, but for this paper, only (a) (ICP sensor) and (b) (iPad mini 4 with MEMS sensor installed on it) specimen will be considered. A series of 12 tests were carried out with varying input amplitude and frequency as shown in <Table 3>. Input frequency ranged from 0.5~ 10.0Hz and input amplitude ranged from 0.01~ 10g. Sampling rate of the iPad mini 4 were set to 100Hz. The time duration for each measurement was 40 seconds.

    4. Shaking-table test results

    Natural frequency and corresponding amplitude for the tests were identified from power spectra and time history curves. Performance of the MEMs sensor was analyzed by comparing the time response curves and power spectra of the two sensors.

    4.1 0.5Hz input frequency

    Two tests (input amplitude of 0.01g and 0.03g) were carried out at this input frequency.

    Time response of the two sensors was compared as shown in <Fig. 7>. The time history response shows results for the first 20 seconds. It can be seen from <Fig. 7> that the MEMS sensor has considerably higher amplitude values than the ICP sensor.

    There is a time lag in time history curve for the MEMS sensor in comparison to the ICP sensor. This might be due to the deviation in sampling frequency. Although the sampling rate was fixed to 100Hz for the iPad device, the actual sampling rate might have changed during the experiment. <Fig. 8> shows the comparison of the power spectra for each of the input amplitudes. The power spectrum was analyzed using all of the data for a time duration of 40 seconds. The corresponding natural frequency values are given.

    4.2 1.0Hz input frequency

    Three tests (input amplitude of 0.01g, 0.03g and 0.1g) were carried out at this input frequency. Time response curves of the two sensors were compared as shown in <Fig. 9>. The time history response shows results for the first 20 seconds. Time lag can also be observed for these test results.

    As explained earlier, the time lag might be due to the deviation in sampling rate of the iPad.

    <Fig. 10> shows the corresponding power spectra. The power spectrum was analyzed using all of the data for a time duration of 40 seconds. MEMS sensor results tend to overestimate the natural frequency values.

    4.3 5.0Hz input frequency

    Five tests (input amplitude of 0.01g, 0.03g, 0.1g, 0.5g and 1.0g) were carried out at this input frequency. Time response and power spectra of the two sensors were compared as shown in <Fig. 11>. For a better visual of the graphs, the x-axis was cut to the first 10 seconds. MEMS sensor tends to overestimate the acceleration values. However, as input acceleration increases, this overestimation reduces. This shows that the MEMS sensor is more accurate at higher input amplitudes.

    <Fig. 12> shows the corresponding response spectra. The power spectrum was analyzed using all of the data for a time duration of 40 seconds.

    4.4 10.0Hz input frequency

    Two tests (input amplitude of 0.5g and 1.0g) were carried out at this input frequency. Time response curves of the two sensors were compared as shown in <Fig. 13>. For a better visual of the graphs, the x-axis was cut to the first 5 seconds. There is no time lag in time history curves for input frequency of 10.0Hz. This shows that accuracy of experiment increases at higher frequencies.

    <Fig. 14> shows the corresponding response spectra. The power spectrum was analyzed using all of the data for a time duration of 40 seconds.

    5. Analysis of shaking-table test results

    5.1 Analysis of error rate in amplitude data

    To analyze the accuracy of the experiment, the error rate (ψa) was calculated using equation (1), where Apeak is the average peak amplitude and Ainput the input amplitude. <Table 4> shows average peak amplitude responses and their corresponding error values.

    ψ a = A p e a k A i n p u t A i n p u t × 100
    (1)

    The range of error values for ICP sensor is from 2.10~17.4% whereas that of the MEMS sensor is 1.04~78.2%. Most of the error values are above 5% error. This shows inaccuracy in obtained amplitude results.

    Lower error values are observed at higher frequencies and the MEMS sensor tends to have lower error values compared to the ICP sensor at these high frequency values(5.0~10.0Hz).

    A general trend of decrease in error as amplitude increases is observed for the MEMS sensor. This is the expected result when using MEMS sensor in vibration measurements.

    To get a clearer image on the deviation of the obtained amplitude values from the input amplitude values, amplitude ratio (peak amplitude/input amplitude) was introduced. <Fig. 15> shows the test results for the amplitude ratio between the input amplitude and the peak amplitude from the sensors. It can be seen that both sensors tend to overestimate the amplitude value for frequencies below 5.0Hz. MEMS sensor at 0.01g has the highest overestimation.

    At higher frequencies (above 5.0Hz), the sensors are more accurate in measuring the amplitude of vibration.

    The response of the two sensors is not accurate in the frequency range of 0.5Hz to 1.0Hz, but the errors of the ICP sensor and the MEMS sensor are relatively compared.

    5.2 Analysis of error rate in frequency data

    To analyze the accuracy of the experiment, the error rate (ψf) was calculated using equation (2).

    ψ f = F n a t u r a l F i n p u t F i n p u t × 100
    (2)

    The natural frequency obtained from the experiment is given as Fnatural with the input frequency being Finput. <Table 5> shows the obtained natural frequency and their corresponding error values. The range of error values is from 0.00~2.60% for the MEMS sensor and 0.00~2.40% for the ICP sensor. All of the error values are below 5% error. This shows accuracy in obtained frequency results.

    However, there are only two anomalous error values (for both sensors at 0.5Hz, 0.03g).

    This might be due to error in measurements during the experiment.

    Lower error values are observed from the power spectrum results compared to the amplitude results. Error values decrease as amplitude and frequency increase.

    To get a clearer image on the deviation of the obtained natural frequency values from the input frequency values, frequency ratio (obtained natural frequency/input frequency) was introduced.

    <Fig. 16> shows the test results for the frequency ratio between the input frequency and the frequency from the sensors. ICP sensor at input amplitude of 0.03g underestimates the value of natural frequency. It can be seen that at higher frequencies (5.0~10.0Hz), the frequency ratio becomes closer to one. This shows that experiments become more accurate at higher frequencies.

    6. Conclusion

    From this experiment, the performance of the Vibration App in obtaining the validity and accuracy of using built-in MEMS sensor on smartphone device was confirmed by a series of shaking-table tests. It was seen that both MEMS and ICP sensors obtained low error values (less than 5%) in natural frequency results whereas they obtained high error values for the amplitude results.

    Although the results of amplitude measurement were not that accurate (errors greater than 5%), the general trend observed for the measurements from the Vibration App was that the error values reduced as amplitude increased. This is the expected behavior of MEMS sensor, which shows that the Vibration App is capable of measuring data from MEMS sensors.

    From the obtained results, it can be concluded that MEMS sensor can be used at higher input frequencies (from 5.0Hz) and at higher input amplitudes (from 0.1g) in obtaining amplitude response of structures. It can also be concluded that MEMS sensors can be used in obtaining natural frequency of structures for both lower and higher input frequencies and amplitudes.

    It can be seen that the performance of the MEMS sensor is generally good, in comparison to the ICP sensor. However, improvements can be made to obtain more accurate results.

    Acknowledgements

    This research was financially supported by the National Research Foundation of Korea (NRF-2016R1A2B2014064). The writers are grateful to the authorities for their support.

    Figure

    KASS-18-49_F1.gif

    iPad mini 4 axial direction9)

    KASS-18-49_F2.gif

    Time series view7)

    KASS-18-49_F3.gif

    Frequency view7)

    KASS-18-49_F4.gif

    Data acquisition settings2)

    KASS-18-49_F5.gif

    Equipments used2)

    KASS-18-49_F6.gif

    Experimental setup2)

    KASS-18-49_F7.gif

    Time history at 0.5Hz for (a) 0.01g and (b) 0.03g

    KASS-18-49_F8.gif

    Power spectrum at 0.5Hz for (a) 0.01g and (b) 0.03g

    KASS-18-49_F9.gif

    Time history at 1.0Hz for (a) 0.01g, (b) 0.03g and (c) 0.1g

    KASS-18-49_F10.gif

    Power spectrum at 1.0Hz for (a) 0.01g, (b) 0.03g and (c) 0.1g

    KASS-18-49_F11.gif

    Time history at 5.0Hz for (a) 0.01g, (b) 0.03g, (c) 0.1g, (d) 0.5g and (e) 1.0g

    KASS-18-49_F12.gif

    Power spectrum at 5.0Hz for (a) 0.01g, (b) 0.03g, (c) 0.1g, (d) 0.5g and (e) 1.0g

    KASS-18-49_F13.gif

    Time history at 10Hz for (a) 0.5g and (b) 1.0g

    KASS-18-49_F14.gif

    Power spectrum at 10Hz for (a) 0.5g and (b) 1.0g

    KASS-18-49_F15.gif

    Peak amplitude-to-input amplitude ratio for (a) MEMS and (b) ICP sensor

    KASS-18-49_F16.gif

    Measured natural frequency-to-input frequency ratio for (a) MEMS and (b) ICP sensor

    Table

    List of laboratory equipments2)

    Characteristics of sensors used2)

    Shaking table experiment overview2)

    Error calculations for amplitude results

    Error calculations for natural frequency results

    Reference

    1. Kareem, A. , Kijewski, T. , & Tamura, Y. , " Mitigation of Motions of Tall Buildings with Specific Examples of Recent Applications ", Wind and Structures, Vol.2,No.3, pp.201~251, 1999
    2. Jung, Y. S. (2017). A study on Low-Noise Vibration Measurement System of the MEMS Sensor (Master's thesis). Seoul National University of Science and Technology
    3. Feng, M. , Fukuda, Y. , Mizuta, M. , & Ozer, E. , " Citizen sensors for SHM: use of accelerometer data from smartphones ", Sensors, Vol.15, No.2, pp.2980~2998, 2015
    4. Kim, J. H. , Min, K. W. , Lee, S. H. , & Moon, D. H. , "Development of monitoring specification of ambient and earthquake responses of building structures", National Disaster Management Research Institute, pp.36~50, 2015
    5. Naito, S. , Azuma, H. , Senna, S. , & Yoshida, M. , " Development and Testing of a Mobile Application for Recording and Analyzing Seismic Data ", Journal of Disaster Research, Vol.8, No.5, pp.990~1000, 2013
    6. Shrestha, A. , Dang, J. , & Wang, X. , " Development of a smart‐device‐based vibration‐measurement system: Effectiveness examination and application cases to existing structure ", Structural Control Health Monitoring, Vol.25, No.3, 2018
    7. Vibration Version 2 [Software]. Diffraction Limited Design LLC. Retrieved from http://www.dld-llc.com
    8. Bosch Sensortec [Apparatus]. (2018). Bosch. Retrieved from https://www.bosch-sensortec.com/
    9. iPad Mini 4 [Apparatus and software]. (2014). Apple Inc. Retrieved from https://web.archive.org/web/20141016183214/http://www.apple.com/ipad-mini-3/