 | Mohammad Hesam Soleimani-Babakamali; Roksana Soleimani-Babakamali; Rodrigo Sarlo A general framework for supervised structural health monitoring and sensor output validation mitigating data imbalance with generative adversarial networks-generated high-dimensional features Journal Article Structural Health Monitoring, 2021. Abstract | Links | BibTeX @article{doi:10.1177/14759217211025488,
title = {A general framework for supervised structural health monitoring and sensor output validation mitigating data imbalance with generative adversarial networks-generated high-dimensional features},
author = {Mohammad Hesam Soleimani-Babakamali and Roksana Soleimani-Babakamali and Rodrigo Sarlo},
doi = {10.1177/14759217211025488},
year = {2021},
date = {2021-07-13},
journal = {Structural Health Monitoring},
abstract = {This study proposes a novelty-classification framework that applies to structural health monitoring (SHM) and sensor output validation (SOV) problems. The proposed framework has simple high-dimensional features with several advantages. First, the feature extraction method is extensively applicable to instrumented structures. Second, the high-dimensional features’ utilization alleviates one of the main issues of supervised novelty classifications, namely, imbalanced datasets and low-sampled data classes. Recurrent Neural Networks are employed for the classification of high-dimensional features. Furthermore, generative adversarial networks (GAN) are trained with low-sampled data classes’ high-dimensional features for generating new data objects. The generated data objects are combined with the initial training set for improving classification results. The proposed framework is studied on two SHM and SOV datasets. The SHM dataset has twenty-one data classes, with a total test accuracy of 99.60% compared to another study with 88.13% accuracy. The SOV classification shows improved results with a mean accuracy of 96.5% compared to three other studies with mean accuracy values of 93.5%, 92.97%, and 71.1%. Furthermore, the integration of GAN’s generated data objects with low-sampled classes improved those classes’ mean F1 score from 44.77% to 64.58% and from 73.39% to 90.84% on SOV and SHM case studies, respectively. The integration of GAN-generated data objects with the initial low-sampled data classes for accuracy improvement shows more potential in the SHM dataset than the SOV case, which can be due to the signal pattern-based labeling logic of SOV datasets.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This study proposes a novelty-classification framework that applies to structural health monitoring (SHM) and sensor output validation (SOV) problems. The proposed framework has simple high-dimensional features with several advantages. First, the feature extraction method is extensively applicable to instrumented structures. Second, the high-dimensional features’ utilization alleviates one of the main issues of supervised novelty classifications, namely, imbalanced datasets and low-sampled data classes. Recurrent Neural Networks are employed for the classification of high-dimensional features. Furthermore, generative adversarial networks (GAN) are trained with low-sampled data classes’ high-dimensional features for generating new data objects. The generated data objects are combined with the initial training set for improving classification results. The proposed framework is studied on two SHM and SOV datasets. The SHM dataset has twenty-one data classes, with a total test accuracy of 99.60% compared to another study with 88.13% accuracy. The SOV classification shows improved results with a mean accuracy of 96.5% compared to three other studies with mean accuracy values of 93.5%, 92.97%, and 71.1%. Furthermore, the integration of GAN’s generated data objects with low-sampled classes improved those classes’ mean F1 score from 44.77% to 64.58% and from 73.39% to 90.84% on SOV and SHM case studies, respectively. The integration of GAN-generated data objects with the initial low-sampled data classes for accuracy improvement shows more potential in the SHM dataset than the SOV case, which can be due to the signal pattern-based labeling logic of SOV datasets. |
 | Mohammad Hesam Soleimani-Babakamali; A Moghadam; Rodrigo Sarlo; M H Hebdon; P S Harvey Mast Arm Monitoring via Traffic Camera Footage: A Pixel-Based Modal Analysis Approach Journal Article Experimental Techniques, 2021, ISBN: 1747-1567. Abstract | Links | BibTeX @article{Soleimani2021,
title = {Mast Arm Monitoring via Traffic Camera Footage: A Pixel-Based Modal Analysis Approach},
author = {Mohammad Hesam Soleimani-Babakamali and A Moghadam and Rodrigo Sarlo and M H Hebdon and P S Harvey},
url = {https://doi.org/10.1007/s40799-020-00422-4},
doi = {10.1007/s40799-020-00422-4},
isbn = {1747-1567},
year = {2021},
date = {2021-01-01},
journal = {Experimental Techniques},
abstract = {Traffic signal mast arm structures must be regularly inspected for cracking, bolt loosening, and other signs of deterioration. Due to large inventories, physical inspections and/or dedicated monitoring systems can be prohibitively time-consuming and expensive to implement at a large scale. However, the growing use of vision-based methods for structural monitoring applications introduces the possibility of leveraging video footage from existing traffic cameras for this purpose. The extraction of dynamic properties (i.e., natural frequencies and damping) from this footage could be employed in detecting possible signs of deterioration. This study presents a vision-based monitoring method which uses a single traffic camera to identify the modal properties of the supporting traffic signal mast arm. This was achieved via operational modal analysis on pixel displacements obtained from a traffic camera mounted on a traffic signal mast arm in Norfolk, VA, monitored during July, 2019. First, sub-pixel displacements were extracted frame-by-frame using weighted centroid tracking of pavement markings. Then, covariance-driven stochastic subspace identification (SSI-Cov) was employed to extract the mast arm fundamental frequencies, damping ratios, and mode shapes. For validation of the vision-based results, SSI-Cov was also applied to acceleration data recorded by two high-sensitivity accelerometers mounted on the structure. In total, the processing was carried out on four different videos and ten acceleration datasets. The vision-based method was able to reliably identify the fundamental frequencies of the structure (Δf < 0.005 Hz mean difference). The associated damping ratios were consistently overestimated but still close in structural terms (Δζ< 0.7% mean difference).},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Traffic signal mast arm structures must be regularly inspected for cracking, bolt loosening, and other signs of deterioration. Due to large inventories, physical inspections and/or dedicated monitoring systems can be prohibitively time-consuming and expensive to implement at a large scale. However, the growing use of vision-based methods for structural monitoring applications introduces the possibility of leveraging video footage from existing traffic cameras for this purpose. The extraction of dynamic properties (i.e., natural frequencies and damping) from this footage could be employed in detecting possible signs of deterioration. This study presents a vision-based monitoring method which uses a single traffic camera to identify the modal properties of the supporting traffic signal mast arm. This was achieved via operational modal analysis on pixel displacements obtained from a traffic camera mounted on a traffic signal mast arm in Norfolk, VA, monitored during July, 2019. First, sub-pixel displacements were extracted frame-by-frame using weighted centroid tracking of pavement markings. Then, covariance-driven stochastic subspace identification (SSI-Cov) was employed to extract the mast arm fundamental frequencies, damping ratios, and mode shapes. For validation of the vision-based results, SSI-Cov was also applied to acceleration data recorded by two high-sensitivity accelerometers mounted on the structure. In total, the processing was carried out on four different videos and ten acceleration datasets. The vision-based method was able to reliably identify the fundamental frequencies of the structure (Δf < 0.005 Hz mean difference). The associated damping ratios were consistently overestimated but still close in structural terms (Δζ< 0.7% mean difference). |