Identification of Full-Field Dynamic Loads on Structures Using Computer Vision and Unsupervised Machine Learning

Roeder, Alexander; Zhang, Huiying; Sánchez, Lorenzo; Yang, Yongchao; Farrar, Charles R.; Mascareñas, David

doi:10.1007/978-3-319-54735-0_5

Cited by 4 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is no need to collect a calibrated displacement measurement from the video. In addition, full-field, high-resolution, unsupervised learning-based structural dynamics identification techniques have been utilized for inferring the load being applied to a structure [41]. If these techniques are made robust for forward deployment then it is possible to envision being able to use them to quantitatively characterize the structural integrity of critical infrastructure.…”

Section: Introductionmentioning

confidence: 99%

Sparse and Random Sampling Techniques for High-Resolution, Full-Field, BSS-Based Structural Dynamics Identification from Video

Martinez

Green

Silva

et al. 2020

Sensors

Self Cite

View full text Add to dashboard Cite

Video-based techniques for identification of structural dynamics have the advantage that they are very inexpensive to deploy compared to conventional accelerometer or strain gauge techniques. When structural dynamics from video is accomplished using full-field, high-resolution analysis techniques utilizing algorithms on the pixel time series such as principal components analysis and solutions to blind source separation the added benefit of high-resolution, full-field modal identification is achieved. An important property of video of vibrating structures is that it is particularly sparse. Typically video of vibrating structures has a dimensionality consisting of many thousands or even millions of pixels and hundreds to thousands of frames. However the motion of the vibrating structure can be described using only a few mode shapes and their associated time series. As a result, emerging techniques for sparse and random sampling such as compressive sensing should be applicable to performing modal identification on video. This work presents how full-field, high-resolution, structural dynamics identification frameworks can be coupled with compressive sampling. The techniques described in this work are demonstrated to be able to recover mode shapes from experimental video of vibrating structures when 70% to 90% of the frames from a video captured in the conventional manner are removed.

show abstract

Section: Introductionmentioning

confidence: 99%

Sparse and Random Sampling Techniques for High-Resolution, Full-Field, BSS-Based Structural Dynamics Identification from Video

Martinez

Green

Silva

et al. 2020

Sensors

Self Cite

View full text Add to dashboard Cite

show abstract

“…In addition, they are computationally extremely extensive. [39][40][41] Lately, phase-based video motion processing technique [42][43][44] has been explored to perform structural dynamics response (displacement) measurements and output-only modal identification without the need for installing markers or speckle paints on a structure's surface, [45][46][47][48][49][50][51] as required by traditional photogrammetry techniques. The recently developed technique enables extraction and visualization of full-field structural dynamic response and modal parameters at high spatial (pixel) resolution locations in an automated manner, 41,51 even with temporally aliased (sub-Nyquist) video measurements avoiding the need of using high-speed digital video cameras.…”

Section: Introductionmentioning

confidence: 99%

Estimation of full‐field dynamic strains from digital video measurements of output‐only beam structures by video motion processing and modal superposition

Yang

Jung

Dorn

et al. 2019

Struct Control Health Monit

Self Cite

View full text Add to dashboard Cite

Summary Strain is an essential quantity to characterize local structural behaviors and directly correlates with structural damage initiation and development that is within local regions. Strain measurement at high spatial resolution (density) locations is thus required to characterize local structural behaviors and detect potential local damage. Traditional contact‐type strain gauges are mostly discrete point‐wise sensors that can only be placed in a limited number of positions. Distributed optical fiber sensing techniques can measure strains at spatially dense measurement points, but their instrumentation is a time‐ and labor‐intensive process associated with the issue of the fragility of fibers. Noncontact optical measurement techniques, such as a family of interferometry techniques using laser beams (e.g., laser Doppler vibrometers), can provide vibration measurement at high density spatial points without the need to install sensors on the structure. However, these measurement devices are active sensing methods that are relatively expensive and vulnerable to ambient motion. Photogrammetry is an alternative noncontact optical measurement method using (passive) white‐light imaging of digital video cameras that are relatively low‐cost, agile, and provides simultaneous measurements at high spatial density locations where every pixel becomes a measurement point. Among others, digital image correlation can achieve full‐field deformation measurements and subsequently estimate the full‐field strains. However, it is computationally extensive. This study develops a new efficient approach to estimate the full‐field (as many measurement points as the pixel number of the video frame on the structure) dynamic strains at high‐spatial (pixel)‐resolution/density location points from the digital video measurement of output‐only vibrating structures. The developed approach is based on phase‐based video motion estimation and modal superposition of structural dynamic response. Furthermore, the method is augmented by a high‐fidelity finite element model, which is updated with the full‐field experimental modal parameters “blindly” identified from the video measurement of the output‐only structure. Laboratory experiments are conducted to validate the method on a bench‐scale cantilever beam structure. Results demonstrate that the full‐field dynamic strain estimated by the developed approach from the video measurement of the output‐only vibrating beam match very well those directly measured by the strain gauges (at discrete measurement points). Some factors associated with the effectiveness of the method are experimentally studied and discussed.

show abstract

“…Using image processing algorithms such as digital image correlation, video camera imaging measurements have been successfully used for vibration measurements of civil structures and more recently for full‐field experimental modal analysis where full‐field modal parameters (mode shapes) can be obtained. Lately, phase‐based video motion processing technique has been explored to perform output‐only modal identification without the need of installing markers or speckle paints on a structure's surface . The latest development of this technique enables extraction and visualization of full‐field, high spatial resolution modal parameters in a relatively efficient and automated manner .…”

Section: Introductionmentioning

confidence: 99%

Estimation of full‐field, full‐order experimental modal model of cable vibration from digital video measurements with physics‐guided unsupervised machine learning and computer vision

Yang

Sánchez

Zhang

et al. 2019

Struct Control Health Monit

Self Cite

View full text Add to dashboard Cite

Cables are critical components for a variety of structures such as stay cables and suspenders of cable-stayed bridges and suspension bridges. When in operational service, they are vulnerable to cumulative fatigue damage induced by dynamic loads (e.g., the cyclic vehicle loads and wind excitation). To accurately analyze and predict their dynamics behaviors and performance that could be spatially local and temporal transient, it is essential to perform highresolution vibration measurements, from which their dynamics properties are identified and, subsequently, a high spatial resolution, full-modal-order dynamics model of cable vibration can be established. This study develops a physics-guided, unsupervised machine learning-based video processing approach that can blindly and efficiently extract the fullfield (as many points as the pixel number of the video frame) modal parameters of cable vibration using only the video of an operating (output-only) cable. In particular, by incorporating the physics of cable vibration (taut string model), a novel automated modal motion filtering method is proposed to enable autonomous identification of full-order (as many modes as possible) dynamic parameters, including those weakly excited modes that used to be challenging to identify in operational modal analysis. Therefore, a full-field, full-order modal model of cable vibration is established by the proposed method. Furthermore, this new approach provides a low-cost and noncontact technique to estimate the cable tension using only the video of the vibrating cable where the fundamental frequency is automatically and efficiently estimated to compute the cable tension according to the taut string equation. Laboratory experiments on a bench-scale cable are conducted to validate the developed approach.

show abstract

Identification of Full-Field Dynamic Loads on Structures Using Computer Vision and Unsupervised Machine Learning

Cited by 4 publications

References 9 publications

Sparse and Random Sampling Techniques for High-Resolution, Full-Field, BSS-Based Structural Dynamics Identification from Video

Sparse and Random Sampling Techniques for High-Resolution, Full-Field, BSS-Based Structural Dynamics Identification from Video

Estimation of full‐field dynamic strains from digital video measurements of output‐only beam structures by video motion processing and modal superposition

Estimation of full‐field, full‐order experimental modal model of cable vibration from digital video measurements with physics‐guided unsupervised machine learning and computer vision

Contact Info

Product

Resources

About