A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Damage Detection in Structural Health Monitoring Using a One-Dimensional Convolutional Neural Network—The Z24 Bridge Case Study
Recently, one of the most significant study areas in civil engineering has been the application of machine learning (ML) to structural damage identification. Traditional statistical pattern recognition methods, such as artificial neural networks (ANNs), are limited to detecting minor damage to bridges. Deep learning algorithms (DLs) extract complicated high-level abstractions, such as data representations, through a hierarchical learning process. Based on comparatively simpler subtractions created at the preceding level of the hierarchy, complex abstractions are learnt at some level. Similar to standard deep learning algorithms, convolutional neural networks (CNNs) are feed-forward ANNs with alternating convolutional and subsampling layers. The main idea is to use the time-series data from the output-only approach as inputs to the one-dimensional convolutional neural network (1DCNN) processing system. A significant benefit is that 1DCNNs have recently been proposed to provide instant monitoring performance at the forefront of structural condition monitoring. The proposed method is validated with the reference data of the Z24 bridge to classify the damage scenario. The results show that the proposed 1DCNN methods exhibit structural defects with excellent accuracy.
Damage Detection in Structural Health Monitoring Using a One-Dimensional Convolutional Neural Network—The Z24 Bridge Case Study
Recently, one of the most significant study areas in civil engineering has been the application of machine learning (ML) to structural damage identification. Traditional statistical pattern recognition methods, such as artificial neural networks (ANNs), are limited to detecting minor damage to bridges. Deep learning algorithms (DLs) extract complicated high-level abstractions, such as data representations, through a hierarchical learning process. Based on comparatively simpler subtractions created at the preceding level of the hierarchy, complex abstractions are learnt at some level. Similar to standard deep learning algorithms, convolutional neural networks (CNNs) are feed-forward ANNs with alternating convolutional and subsampling layers. The main idea is to use the time-series data from the output-only approach as inputs to the one-dimensional convolutional neural network (1DCNN) processing system. A significant benefit is that 1DCNNs have recently been proposed to provide instant monitoring performance at the forefront of structural condition monitoring. The proposed method is validated with the reference data of the Z24 bridge to classify the damage scenario. The results show that the proposed 1DCNN methods exhibit structural defects with excellent accuracy.
Damage Detection in Structural Health Monitoring Using a One-Dimensional Convolutional Neural Network—The Z24 Bridge Case Study
Lecture Notes in Civil Engineering
Nguyen-Xuan, Tung (editor) / Nguyen-Viet, Thanh (editor) / Bui-Tien, Thanh (editor) / Nguyen-Quang, Tuan (editor) / De Roeck, Guido (editor) / Nguyen-Tran, Hieu (author) / Bui-Ngoc, Dung (author) / Pham-Tuan, Dung (author) / Ngoc-Nguyen, Lan (author) / Tran-Ngoc, Hoa (author)
International Conference on Sustainability in Civil Engineering ; 2022 ; Hanoi, Vietnam
Proceedings of the 4th International Conference on Sustainability in Civil Engineering ; Chapter: 70 ; 683-692
2023-08-13
10 pages
Article/Chapter (Book)
Electronic Resource
English
Structural Damage Detection using Deep Convolutional Neural Network and Transfer Learning
| Springer Verlag | 2019
Structural Damage Detection using Deep Convolutional Neural Network and Transfer Learning
| Online Contents | 2019