Convenient Structural Modal Analysis Using Noncontact Vision-Based Displacement Sensor

Transcription

1 8th European Workshop On Structural Health Monitoring (EWSHM 2016), 5-8 July 2016, Spain, Bilbao Convenient Structural Modal Analysis Using Noncontact Vision-Based Displacement Sensor More info about this article: Dongming FENG 1, Maria Q. FENG 1 1 Department of Civil Engineering and Engineering Mechanics, Columbia University, New York, NY df2465@columbia.edu Key words: Vision sensor, Displacement measurement, Modal analysis, Frame structure, Simply supported beam, Low-cost. Abstract As an emerging noncontact method, the vision-based displacement sensor systems have offered a promising alternative to the conventional sensors for vibration-based structural health monitoring (SHM). While most existing studies on vision sensors focus on the measurement performance evaluation, the potentials of vision sensor for low-cost structural modal analysis are experimentally explored through a three-story frame structure and a simply beam structure. The identified modal results based on the displacement measurements by using one low-cost camera agree well with that from multiple reference accelerometers. It is believed that identified modal parameters can be a better substitute for model updating, system identification, and detect damages, etc., as the vision sensor can achieve smoother mode shapes while the resolution of mode shapes from accelerometers is limited by the sensor number. 1 INTRODUCTION Over the last few decades, considerable efforts have been made toward vibration-based SHM techniques [1-10]. SHM can be defined as a process of damage detection and localization which involves the expertise of sensor technology, data processing, structural modeling, identification algorithm development, etc. The basic principle is that any structural damage or degradation would result in changes in structural dynamic responses and the corresponding modal characteristics. Doebling et al. [5] presented an extensive literature review concerning the detection, location and characterization of structural damage based on changes in the frequency-domain modal properties, such as modal frequencies, mode shapes and its curvatures, modal flexibility coefficients, etc. Currently, most of the existing experimental modal analysis for modal information extraction are based on acceleration records, the measurement of which requires mounting an array of accelerometers on a structure. One main bottleneck lies in that the spatial resolution of the obtained mode shape depends on the total number of deployed point-wise sensors, which may result in less accurate damage localization. In addition, considerable measurement time and effort are required if wired accelerometers are employed. Recently, novel video-based noncontact displacement measurement techniques have been developed, and offer promising alternatives to the conventional contact-type point-wise sensors [11-16]. It can achieve a measurement accuracy less than 0.1mm for both lab and field tests. However, most of the existing vision sensor studies have still focused on the sensor performance evaluation, without discussing the use of the measured displacement

2 data. Only a few attempts have been made towards applying this novel sensor for structural modal analysis, system identification and damage detection, etc. [14, 17-20]. Compared with accelerometers, obtaining mode shapes by vision sensor is expected to be more convenient and more accurate since the multi-point dense displacement measurements from one camera or multiple synchronized cameras would give smoother mode shapes. As an effort to explore the potentials of the vision sensor for low-cost structural modal analysis, in this study, experimental studies are conducted on a scaled three-story frame structure and a simply supported beam specimen, respectively. Moreover, identified modal results are compared with those from reference sensors. 2 VISION SENSOR PRINCIPLE The underlying principle of the vision sensor for displacement measurement is the template matching technique. In the implementation, an initial area is defined as a template in the first image of a sequence of video frames, as illustrated in Figure 1. The template can be located in the successive images using the template matching technique. To reduce computational time, the searching area would be confined to a predefined region of interest (ROI) near the template s location in previous image. Figure 1: Schematic representation of the template matching In this study, based on the robust orientation code matching (OCM) template matching algorithm and subpixel bilinear interpolation, the vision sensor software is developed to accurately track structural displacements at user-defined locations frame-by-frame. Detailed description of the subpixel OCM principle can be found in literature [19]. 3 THREE-STORY FRAME STRUCTURE: SHAKING TABLE TEST The performance of the proposed vision sensor for structural modal analysis is firstly evaluated through a shaking table test of a scaled three-story frame structure in the Carleton Laboratory at Columbia University, as shown in Figure 2. The aluminum frame structure is bolt-connected for all the column-floor connections. During test, the shaking table is driven by generated white noise signal. The visions senor system is placed at a stationary position, 8 meters away from the shaking table. As shown in Figure 2, four high-contrast artificial targets are mounted on the structure for motion tracking. Meanwhile, four bolt connections are used to study the performance of the vision sensor to track low-contrast natural targets on the structure. As references, the displacements are also measured by four high-accuracy laser displacement sensors (LDSs), which are installed between each floor of the frame model and stationary reference points. Besides, four accelerometers are installed to each floor to further compare the experimental modal analysis results. To evaluate the time-domain performance of the vision sensor, displacements are measured by tracking both high-contrast artificial targets and low-contrast natural targets 2

3 (i.e., bolt connections) and compared with those by LDSs. The measurements are respectively termed as Vision (artificial), Vision (natural), and LDS. Figure 2: Shaking table test setup 3.1 Measurement by tracking artificial targets and natural targets Figure 3: Comparison of displacements by Vision (artificial) and LDS: a) Base displacement, b) 3rd floor relative displacement The four artificial targets in Figure 2 are first used as the tracking target for the vision sensor. Figure 3(a) plots the displacement time histories of the base floor by vision sensor and LDS, respectively. Excellent agreements are observed. In Figure 3(b), the plotted displacement of the 3 rd floor is relative to the base displacement shown in Figure 3(a), and only enlarged time segments between 1s to 5.5s are presented for better illustration. To save space, measurements for the 1 st and 2 nd floors are not plotted. Then, instead of using artificial targets, the existing natural targets, i.e., the four bolt connections on the frame structure as shown in Figure 2, are used as the tracking targets. Results in Figure 4 indicates that good agreements between vision sensor and LDS are also observed. It is noted that by tracking the natural target without requiring an artificial target installed on a fixed location of the structure, the vision sensor provides the flexibility to easily change locations for displacement measurement, thus further facilitating the testing 3

4 process. Figure 4; Comparison of displacements by Vision (natural) and LDS: a) Base displacement, b) 3rd floor relative displacement 3.2 Modal analysis Modal analysis is carried out by using the stochastic subspace identification (SSI) method. Table 1 compares the identified natural frequencies from displacement measurements by Vision (artificial), Vision (natural), and from accelerations by accelerometers, respectively. It is observed that the frequency differences between the vision sensor and accelerometer are very small. Natural Freq. Vision (artificial) (Hz) Vision (natural) (Hz) Accelerometer (Hz) 1 st nd rd Table 1: Comparison of identified natural frequencies Figure 5: Comparison of mode shapes: (a) Accelerometer vs. Vision (artificial), and (b) Accelerometer vs. Vision (natural) Figure 5(a) compares the three mode shapes by Vision (artificial) and accelerometers, and Figure 5(b) compares the mode shapes by Vision (natural) and accelerometers. They all match well with one another. It is noted the mode shape is scaled to 1 with respect to a 4

5 reference degree of freedom (herein, 3rd floor). Compared with accelerometers, the vision sensor system is more convenient and cost-efficient since modal analysis can be easily enabled by the multi-point displacements remotely measured from one camera. 4 SIMPLY SUPPORTED BEAM: HAMMER IMPACT TEST As shown in Figure 6, the simply supported beam model is made from rectangular aluminum sheet. 30 predesigned black dots, numbered from 2 through 31, are attached along the beam as targets for motion tracking. As references, six accelerometers are installed to further compare the modal analysis results. Herein, hammer impact is used to excite higher structural vibrational components. The time histories at 30 points along the beam for hammer hit at point 4 are shown in Figure 7. Figure 6: Test setup for the simply supported beam Figure 7: Displacement measurements at point 2 through 31 by vision sensor Figure 18: Frequency results from: (a) displacement measurements at point 2 through 31 by vision sensor, (b) accelerations at 6 points by accelerometers. Figure 8 compares the extracted natural frequencies by vision sensor and accelerometers, respectively, which shows a perfect match. Subsequently, the mode shapes of the simply 5

6 supported beam model can be obtained from modal analysis. Figure 9 compares the first two mode shapes (scaled to 1) by accelerometers and vision sensor. It is indicated that the vision sensor can achieve smoother mode shapes while the resolution of mode shapes from accelerometers is limited by the sensor number. Figure 9: Comparison of mode shapes between vision sensor and accelerometer: (a) 1st mode shape, (b) 2nd mode shape 5. CONCLUSION While most existing studies regarding vision displacement sensors have been focusing on the measurement accuracy evaluation, this study validates the usefulness of the vision sensor for structural modal analysis. The observations of this study can be summarized as follows: (1) From the shaking table test of a frame structure, excellent agreements are observed between the multi-point displacement time histories by the vision sensor by tracking either high-contrast artificial targets or low-contrast natural targets on the structure and those by reference laser displacement sensors. Moreover, modal analysis shows that the obtained natural frequencies and mode shapes from measurements by using one camera match well with those by using four accelerometers. (2) Test results of a simply-supported beam structures further demonstrate that dynamic displacement responses at 30 points can be simultaneously and accurately measured using one camera, and the identified natural frequencies and mode shapes by the vision sensor match well with those by using six accelerometers. (3) It should be pointed out that the vision sensor can achieve smoother mode shapes while the resolution of mode shapes from accelerometers is limited by the sensor number. Thus it can be concluded that the vision sensor is a low-cost high-performance alternative to the conventional accelerometers for structural modal analysis. In practical applications for full-scale structures such as high-rise buildings or long-span bridges, multiple synchronized cameras can be used to tracking displacements at different sections of the structure. REFERENCES [1] Masri S, Smyth A, Chassiakos A, Caughey T, Hunter N. Application of Neural Networks for Detection of Changes in Nonlinear Systems. Journal of Engineering Mechanics. 2000;126: [2] Lee JJ, Lee JW, Yi JH, Yun CB, Jung HY. Neural networks-based damage detection for 6

7 bridges considering errors in baseline finite element models. Journal of Sound and Vibration. 2005;280: [3] Farrar CR, Worden K. An introduction to structural health monitoring. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences. 2007;365: [4] Farrar CR, Doebling SW, Nix DA. Vibration based structural damage identification. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences. 2001;359: [5] Doebling SW, Farrar CR, Prime MB, Shevitz DW. Damage identification and health monitoring of structural and mechanical systems from changes in their vibration characteristics: A literature review. Other Information: PBD: May p. Medium: ED; Size: 132 p. [6] Balsamo L, Betti R. Data-based structural health monitoring using small training data sets. Structural Control and Health Monitoring. 2015;22: [7] Mukhopadhyay S, Luş H, Betti R. Probabilistic Structural Health Assessment with Identified Physical Parameters from Incomplete Measurements. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering. 2015:B [8] Mukhopadhyay S, Lus H, Betti R. Structural identification with incomplete instrumentation and global identifiability requirements under base excitation. Structural Control and Health Monitoring. 2015;22: [9] Feng D, Sun H, Feng MQ. Simultaneous identification of bridge structural parameters and vehicle loads. Computers & Structures. 2015;157: [10] Sun H, Feng D, Liu Y, Feng MQ. Statistical Regularization for Identification of Structural Parameters and External Loadings Using State Space Models. Computer- Aided Civil and Infrastructure Engineering. 2015;30: [11] Fukuda Y, Feng MQ, Shinozuka M. Cost-effective vision-based system for monitoring dynamic response of civil engineering structures. Structural Control and Health Monitoring. 2010;17: [12] Fukuda Y, Feng MQ, Narita Y, Kaneko S, Tanaka T. Vision-Based Displacement Sensor for Monitoring Dynamic Response Using Robust Object Search Algorithm. Sensors Journal, IEEE. 2013;13: [13] Pan B, Qian K, Xie H, Asundi A. Two-dimensional digital image correlation for inplane displacement and strain measurement: a review. Measurement Science and Technology. 2009;20: [14] Song Y-Z, Bowen CR, Kim AH, Nassehi A, Padget J, Gathercole N. Virtual visual sensors and their application in structural health monitoring. Structural Health Monitoring. 2014;13: [15] Feng M, Fukuda Y, Feng D, Mizuta M. Nontarget Vision Sensor for Remote Measurement of Bridge Dynamic Response. Journal of Bridge Engineering. 2015: [16] Feng D, Feng M, Ozer E, Fukuda Y. A Vision-Based Sensor for Noncontact Structural Displacement Measurement. Sensors. 2015;15: [17] Feng D, Feng M. Model Updating of Railway Bridge Using In Situ Dynamic Displacement Measurement under Trainloads. Journal of Bridge Engineering. 2015: [18] Dworakowski Z, Kohut P, Gallina A, Holak K, Uhl T. Vision-based algorithms for damage detection and localization in structural health monitoring. Structural Control and 7

8 Health Monitoring doi: /stc [19] Feng D, Feng MQ. Vision-based multipoint displacement measurement for structural health monitoring. Structural Control and Health Monitoring. 2015: [20] Feng DM. Advanced vision-based displacement sensors for structural health monitoring. Columbia University Academic Commons