Ge Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ge Zhang has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Fangsen Cui
Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), 1 Fusionopolis Way, #16-16 Connexis, Singapore 138632, Singapore
3 shared publications - Gongfa Chen
School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006, P. R. China
3 shared publications - Bao Yang
Department of Engineering Mechanics, School of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510640, China
2 shared publications - Hua Liu
Guangzhou Power Supply Bureau, Guangdong Power Grid Co., Ltd., China Southern Power Grid Co., Ltd., Guangzhou 510665, China
2 shared publications Hui Sun
2 shared publications- Yongjian Cai
School of Civil Engineering and Transportation, Foshan University, Foshan 528225, China
2 shared publications
Latest Publications
- Open Access
- 19 November 2026
- Published by Tsinghua University Press in Lifeline Emergency and Safety
- Vol. 1 (4)
Prediction on tensile mechanical properties of engineered cementitious composites using a fully automated deep learning method
Accurately predicting tensile mechanical properties of Engineered Cementitious Composites (ECC), including tensile strength (TS) and ultimate tensile strain (UTS), remains challenging, particularly for the ultimate tensile strain, which exhibits high data dispersion that hinders reliable predictions. While machine learning (ML) has shown significant promise in predicting the mechanical properties of ECC, it faces several challenges: (1) limited predictive accuracy; (2) complexity and time-consuming nature of hyperparameter tuning; (3) insufficient research into predicting the UTS of ECC. To overcome these challenges, this study proposes a fully automated deep learning model enhanced by Bayesian optimization for predicting the tensile mechanical properties of polyethylene fiber-reinforced engineered cementitious composites (PE-ECC). First, convolutional neural network (CNN) model was employed to enhance feature extraction and predictive accuracy. Second, Bayesian optimization was utilized to automatically construct the network and optimize various hyperparameters, resulting in an optimized network. This research demonstrates the effectiveness of the proposed Bayesian-optimized convolutional neural network (BO-CNN) method in predicting PE-ECC’s TS and UTS, utilizing a dataset of 154 experimental samples from published literature, achieving coefficients of determination ( R2) of 0.911 and 0.775, respectively. The BO-CNN method not only outperforms CNNs and artificial neural networks (ANNs) optimized through grid search but also surpasses other eleven traditional ML models. Furthermore, SHapley Additive exPlanations (SHAP) was employed to interpret input features’ contributions in the prediction of tensile mechanical properties. Overall, the proposed method provides an efficient and reliable solution for predicting the tensile mechanical properties of ECC.
- 19 November 2026
- Published by Tsinghua University Press in Lifeline Emergency and Safety
- Vol. 1 (4)
Open AccessPrediction on tensile mechanical properties of engineered cementitious composites using a fully automated deep learning method
Accurately predicting tensile mechanical properties of Engineered Cementitious Composites (ECC), including tensile strength (TS) and ultimate tensile strain (UTS), remains challenging, particularly for the ultimate tensile strain, which exhibits high data dispersion that hinders reliable predictions. While machine learning (ML) has shown significant promise in predicting the mechanical properties of ECC, it faces several challenges: (1) limited predictive accuracy; (2) complexity and time-consuming nature of hyperparameter tuning; (3) insufficient research into predicting the UTS of ECC. To overcome these challenges, this study proposes a fully automated deep learning model enhanced by Bayesian optimization for predicting the tensile mechanical properties of polyethylene fiber-reinforced engineered cementitious composites (PE-ECC). First, convolutional neural network (CNN) model was employed to enhance feature extraction and predictive accuracy. Second, Bayesian optimization was utilized to automatically construct the network and optimize various hyperparameters, resulting in an optimized network. This research demonstrates the effectiveness of the proposed Bayesian-optimized convolutional neural network (BO-CNN) method in predicting PE-ECC’s TS and UTS, utilizing a dataset of 154 experimental samples from published literature, achieving coefficients of determination ( R2) of 0.911 and 0.775, respectively. The BO-CNN method not only outperforms CNNs and artificial neural networks (ANNs) optimized through grid search but also surpasses other eleven traditional ML models. Furthermore, SHapley Additive exPlanations (SHAP) was employed to interpret input features’ contributions in the prediction of tensile mechanical properties. Overall, the proposed method provides an efficient and reliable solution for predicting the tensile mechanical properties of ECC.
- Open Access
Integrating Transformer Segmentation and Morphological Analysis for Multi-Crack Characterization and Visual Equivalent Strain Assessment in Rubberized ECC
Deep-learning semantic segmentation and morphological analysis offer a non-contact route to characterizing microcracks in cementitious composites, yet linking pixel-level masks to physical geometry and tensile response remains challenging. This study establishes an integrated lightweight segmentation and geometric-quantification workflow for rubberized engineered cementitious composites (ECCs) incorporating 0–50% powdered rubber. We propose EfficientSegLite, a supervised binary segmentation architecture combining a hierarchical SegFormer-B0 encoder, a channel-compressed U-Net-style decoder, lightweight atrous spatial pyramid pooling, and depthwise separable convolutions, trained via a two-stage Dice–OHEM–Edge strategy. On 1440 ECC tensile images, it achieved 93.40% mIoU and 93.24% F1 score with only 3.96 M parameters and 734.9 M FLOPs. Deterministic morphological and distance-transform algorithms then convert the masks into calibrated physical descriptors, achieving a mean relative crack-width error of 3.65% against optical microscopy across eight benchmark cracks. A crack-area-based visual equivalent strain (ε_v) correlated strongly with extensometer strain (Pearson’s r > 0.97, R2 > 0.94); however, regression slopes of 0.368–1.046 and uncalibrated mean absolute errors of 0.03–2.04 percentage points indicate that quantitative strain estimation requires mixture-specific calibration. A tool-calling large language model agent parses user instructions and orchestrates the deterministic computation tools, generating traceable inspection reports.
Open AccessIntegrating Transformer Segmentation and Morphological Analysis for Multi-Crack Characterization and Visual Equivalent Strain Assessment in Rubberized ECC
Deep-learning semantic segmentation and morphological analysis offer a non-contact route to characterizing microcracks in cementitious composites, yet linking pixel-level masks to physical geometry and tensile response remains challenging. This study establishes an integrated lightweight segmentation and geometric-quantification workflow for rubberized engineered cementitious composites (ECCs) incorporating 0–50% powdered rubber. We propose EfficientSegLite, a supervised binary segmentation architecture combining a hierarchical SegFormer-B0 encoder, a channel-compressed U-Net-style decoder, lightweight atrous spatial pyramid pooling, and depthwise separable convolutions, trained via a two-stage Dice–OHEM–Edge strategy. On 1440 ECC tensile images, it achieved 93.40% mIoU and 93.24% F1 score with only 3.96 M parameters and 734.9 M FLOPs. Deterministic morphological and distance-transform algorithms then convert the masks into calibrated physical descriptors, achieving a mean relative crack-width error of 3.65% against optical microscopy across eight benchmark cracks. A crack-area-based visual equivalent strain (ε_v) correlated strongly with extensometer strain (Pearson’s r > 0.97, R2 > 0.94); however, regression slopes of 0.368–1.046 and uncalibrated mean absolute errors of 0.03–2.04 percentage points indicate that quantitative strain estimation requires mixture-specific calibration. A tool-calling large language model agent parses user instructions and orchestrates the deterministic computation tools, generating traceable inspection reports.
- 19 September 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Applied Mechanics
- Vol. 18 (11)
A Cable Force Anomaly Identification Method Based on Bridge Deck Vibration Signals and an Unsupervised Convolutional Autoencoder
Supervised cable-force anomaly detection requires labeled damage data that are rarely available for in-service bridges. This study therefore proposes an unsupervised convolutional autoencoder (CAE) that identifies cable-force anomalies from bridge-deck vibration signals. Finite-element simulations provide training data, and the CAE learns only healthy-state vibration patterns. Reconstruction quality is evaluated using three complementary indices: reconstruction error (ERROR), structural similarity (SSIM), and the original-to-reconstructed signal ratio (ORSR). Extreme value theory provides reproducible unsupervised thresholds, while clustering reveals the progression of cable-force loss. The method was validated using a scaled model of the First Zhongshan Bridge and a full-scale cable-stayed bridge. The three indices distinguished normal and abnormal states and exhibited progressive changes with increasing cable-force loss, demonstrating the feasibility of deck-vibration-based cable-force anomaly identification.
- 19 September 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Applied Mechanics
- Vol. 18 (11)
A Cable Force Anomaly Identification Method Based on Bridge Deck Vibration Signals and an Unsupervised Convolutional Autoencoder
Supervised cable-force anomaly detection requires labeled damage data that are rarely available for in-service bridges. This study therefore proposes an unsupervised convolutional autoencoder (CAE) that identifies cable-force anomalies from bridge-deck vibration signals. Finite-element simulations provide training data, and the CAE learns only healthy-state vibration patterns. Reconstruction quality is evaluated using three complementary indices: reconstruction error (ERROR), structural similarity (SSIM), and the original-to-reconstructed signal ratio (ORSR). Extreme value theory provides reproducible unsupervised thresholds, while clustering reveals the progression of cable-force loss. The method was validated using a scaled model of the First Zhongshan Bridge and a full-scale cable-stayed bridge. The three indices distinguished normal and abnormal states and exhibited progressive changes with increasing cable-force loss, demonstrating the feasibility of deck-vibration-based cable-force anomaly identification.
- Open Access
- 8 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (16) , 7912
Deep Learning-Based Cross-Verification for Road Subsurface Distress Detection Driven by Field Data of 3D Ground-Penetrating Radar
Ground-penetrating radar (GPR) is a rapid and non-destructive technique for road sub-surface distress (RSD) detection. However, reliable interpretation of GPR images remains challenging because subsurface anomalies often present weak boundaries, ambiguous textures, and high similarity to non-distress targets. This study proposes a cross-verification intelligent algorithm that exploits complementary information from different views of 3D GPR data. Three YOLO-based detectors are trained on view-specific GPR images to identify RSD-related targets, including voids, loose structures, and manholes. By sequentially verifying detection results across different views, the proposed method improves recognition reliability under vague subsurface imaging conditions. The models are trained and evaluated on an expert-annotated field 3D GPR dataset containing 2134 location-level multi-view samples. At the selected operational thresholds, the complete cross-verification procedure achieved 95.9% precision and 98.6% recall for RSD detection in the testing subset. In a field evaluation on 15 roads, all 69 RSD locations in the expert-identified reference set were matched by automatic indications. When integrated into an automatic detection system, the method reduced manual inspection workloads by approximately 90% while maintaining high field reliability. These results demonstrate the potential of multi-view cross-verification for post-survey RSD screening and expert-assisted review.
- 8 August 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (16) , 7912
Open AccessDeep Learning-Based Cross-Verification for Road Subsurface Distress Detection Driven by Field Data of 3D Ground-Penetrating Radar
Ground-penetrating radar (GPR) is a rapid and non-destructive technique for road sub-surface distress (RSD) detection. However, reliable interpretation of GPR images remains challenging because subsurface anomalies often present weak boundaries, ambiguous textures, and high similarity to non-distress targets. This study proposes a cross-verification intelligent algorithm that exploits complementary information from different views of 3D GPR data. Three YOLO-based detectors are trained on view-specific GPR images to identify RSD-related targets, including voids, loose structures, and manholes. By sequentially verifying detection results across different views, the proposed method improves recognition reliability under vague subsurface imaging conditions. The models are trained and evaluated on an expert-annotated field 3D GPR dataset containing 2134 location-level multi-view samples. At the selected operational thresholds, the complete cross-verification procedure achieved 95.9% precision and 98.6% recall for RSD detection in the testing subset. In a field evaluation on 15 roads, all 69 RSD locations in the expert-identified reference set were matched by automatic indications. When integrated into an automatic detection system, the method reduced manual inspection workloads by approximately 90% while maintaining high field reliability. These results demonstrate the potential of multi-view cross-verification for post-survey RSD screening and expert-assisted review.
- Open Access
- 31 July 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (15) , 7616
Integrated UAV Path Planning and Attention-Enhanced Instance Segmentation for Automated Infrastructure Surface Defect Detection
This paper presents an integrated framework for automated detection of surface defects in infrastructures using unmanned aerial vehicles (UAVs), comprising 3D model-based adaptive path planning, high-resolution image acquisition, and an attention-enhanced instance segmentation model. However, existing approaches face two key limitations: (i) conventional UAV path planning lacks adaptive trajectory correction for non-horizontal bridge geometries, and (ii) instance segmentation models for infrastructure defects have not been systematically optimized for both accuracy and edge-device deployability. To address these gaps, the proposed framework was trained on 3625 annotated images covering two defect categories (spalling and cracking) and preliminarily validated through a proof-of-concept field study on a concrete viaduct section where seven spalling and one reinforcement exposure were detected. The methodology consists of three core components: (i) 3D model-based adaptive path planning, (ii) high-resolution image acquisition under variable infrastructure geometries, and (iii) an improved instance segmentation model based on YOLOv8-seg. To ensure consistent imaging geometry, a segmented linear interpolation method is introduced to adaptively correct flight trajectories for non-horizontal infrastructure sections. For damage detection, we propose a structurally enhanced YOLOv8-seg model, denoted as YOLOv8-seg-ECAC2f-all, which integrates Efficient Channel Attention (ECA) modules into a fully modified backbone architecture. Compared to the baseline YOLOv8-seg, the proposed model achieves a mean average precision (mAP50–95) of 91.8% for bounding box detection and 61.7% for instance segmentation, corresponding to improvements of 7.7% and 2.8%, respectively. The framework was preliminarily validated on a concrete viaduct of the Guangfojiangzhu Expressway, achieving 100% inspection coverage and detection of all eight ground-truth surface defects (seven spalling and one reinforcement exposure) in this pilot study, including two minor spalling cases (≤0.5 m2) missed by manual inspection. These results demonstrate the technical feasibility of the proposed framework for real-world concrete bridge inspection and its potential for reducing manual inspection effort while improving detection sensitivity for minor defects ().- 31 July 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (15) , 7616
Open AccessIntegrated UAV Path Planning and Attention-Enhanced Instance Segmentation for Automated Infrastructure Surface Defect Detection
This paper presents an integrated framework for automated detection of surface defects in infrastructures using unmanned aerial vehicles (UAVs), comprising 3D model-based adaptive path planning, high-resolution image acquisition, and an attention-enhanced instance segmentation model. However, existing approaches face two key limitations: (i) conventional UAV path planning lacks adaptive trajectory correction for non-horizontal bridge geometries, and (ii) instance segmentation models for infrastructure defects have not been systematically optimized for both accuracy and edge-device deployability. To address these gaps, the proposed framework was trained on 3625 annotated images covering two defect categories (spalling and cracking) and preliminarily validated through a proof-of-concept field study on a concrete viaduct section where seven spalling and one reinforcement exposure were detected. The methodology consists of three core components: (i) 3D model-based adaptive path planning, (ii) high-resolution image acquisition under variable infrastructure geometries, and (iii) an improved instance segmentation model based on YOLOv8-seg. To ensure consistent imaging geometry, a segmented linear interpolation method is introduced to adaptively correct flight trajectories for non-horizontal infrastructure sections. For damage detection, we propose a structurally enhanced YOLOv8-seg model, denoted as YOLOv8-seg-ECAC2f-all, which integrates Efficient Channel Attention (ECA) modules into a fully modified backbone architecture. Compared to the baseline YOLOv8-seg, the proposed model achieves a mean average precision (mAP50–95) of 91.8% for bounding box detection and 61.7% for instance segmentation, corresponding to improvements of 7.7% and 2.8%, respectively. The framework was preliminarily validated on a concrete viaduct of the Guangfojiangzhu Expressway, achieving 100% inspection coverage and detection of all eight ground-truth surface defects (seven spalling and one reinforcement exposure) in this pilot study, including two minor spalling cases (≤0.5 m2) missed by manual inspection. These results demonstrate the technical feasibility of the proposed framework for real-world concrete bridge inspection and its potential for reducing manual inspection effort while improving detection sensitivity for minor defects (). - Open Access
- 27 July 2026
- Published by Elsevier in Case Studies in Construction Materials
- Vol. 25
A sample efficient deep learning framework for bond performance assessment of FRP bar reinforced seawater sea sand concrete
- 27 July 2026
- Published by Elsevier in Case Studies in Construction Materials
- Vol. 25
Open AccessA sample efficient deep learning framework for bond performance assessment of FRP bar reinforced seawater sea sand concrete
- Open Access
An Intelligent Method for Bearing Pad Flatness Inspection Based on UAV-Enabled 3D Reconstruction
The flatness of bearing pads directly affects structural load transfer safety. However, conventional total station-based inspection methods suffer from limited spatial sampling, low inspection efficiency, and high safety risks associated with working at height. To address these limitations, this paper proposes UAV-FIBP (Unmanned Aerial Vehicle-based Flatness Inspection for Bridge Pads), an automated and intelligent method for pad flatness assessment utilizing UAV-based 3D reconstruction. By designing a close-range, multi-orbit circumnavigational UAV flight path and acquiring high-overlap imagery (85% forward and 80% side overlap), a millimeter-accurate 3D model is generated via photogrammetry, achieving a high-density point cloud of ≥200 points/cm2 on the pad surface. Following point cloud denoising and region-of-interest segmentation using the Random Sample Consensus (RANSAC) algorithm, Principal Component Analysis (PCA) is employed to fit a reference plane. A dual-parameter evaluation framework is subsequently introduced: the root mean square (RMS) deviation quantifies local surface roughness, while the maximum elevation difference is derived from the angle between the normal vectors of the fitted plane and the horizontal plane, thereby enabling a comprehensive assessment of global inclination. Validation experiments conducted on laboratory-scale setups and real construction sites (involving four bridge pads) demonstrate that the proposed method achieves deviations ≤ 2 mm compared to total station measurements, satisfying the requirements stipulated in the Standards for Quality Inspection and Verification of Highways (JTG F80/1-2017). Results indicate that UAV-FIBP enables non-contact, full-coverage, and automated flatness inspection, significantly improving inspection efficiency and construction safety. This work establishes a scalable technical pathway for intelligent bridge construction.
Open AccessAn Intelligent Method for Bearing Pad Flatness Inspection Based on UAV-Enabled 3D Reconstruction
The flatness of bearing pads directly affects structural load transfer safety. However, conventional total station-based inspection methods suffer from limited spatial sampling, low inspection efficiency, and high safety risks associated with working at height. To address these limitations, this paper proposes UAV-FIBP (Unmanned Aerial Vehicle-based Flatness Inspection for Bridge Pads), an automated and intelligent method for pad flatness assessment utilizing UAV-based 3D reconstruction. By designing a close-range, multi-orbit circumnavigational UAV flight path and acquiring high-overlap imagery (85% forward and 80% side overlap), a millimeter-accurate 3D model is generated via photogrammetry, achieving a high-density point cloud of ≥200 points/cm2 on the pad surface. Following point cloud denoising and region-of-interest segmentation using the Random Sample Consensus (RANSAC) algorithm, Principal Component Analysis (PCA) is employed to fit a reference plane. A dual-parameter evaluation framework is subsequently introduced: the root mean square (RMS) deviation quantifies local surface roughness, while the maximum elevation difference is derived from the angle between the normal vectors of the fitted plane and the horizontal plane, thereby enabling a comprehensive assessment of global inclination. Validation experiments conducted on laboratory-scale setups and real construction sites (involving four bridge pads) demonstrate that the proposed method achieves deviations ≤ 2 mm compared to total station measurements, satisfying the requirements stipulated in the Standards for Quality Inspection and Verification of Highways (JTG F80/1-2017). Results indicate that UAV-FIBP enables non-contact, full-coverage, and automated flatness inspection, significantly improving inspection efficiency and construction safety. This work establishes a scalable technical pathway for intelligent bridge construction.
- Open Access
- 22 April 2026
- Published by Elsevier in SSRN Electronic Journal
Automatic Road Subsurface Distress Recognition from Ground Penetrating Radar Images using Deep Learning-based Cross-verification
Ground penetrating radar (GPR) has become a rapid and non-destructive solution for road subsurface distress (RSD) detection. However, recognizing RSD from GPR images is labor-intensive and heavily relies on the expertise of inspectors. Deep learning-based automatic RSD recognition, though ameliorating the burden of data processing, suffers from insufficient capability to recognize defects. In this study, a novel cross-verification strategy was proposed to fully exploit the complementary abilities of region proposal networks in object recognition from different views of GPR images. Following this strategy, three YOLO-based models were used to detect the RSD (voids and loose structures) and manholes. Each model was trained with a specific view of 3D GPR dataset, which contains rigorously validated 2134 samples of diverse types obtained through field scanning. The cross-verification strategy achieves outstanding accuracy with a recall of over 98.6% in the tests using real field-scanning data. Field tests also show that deep learning-based automatic RSD recognition can reduce the human labor of inspection by around 90%.
- 22 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAutomatic Road Subsurface Distress Recognition from Ground Penetrating Radar Images using Deep Learning-based Cross-verification
Ground penetrating radar (GPR) has become a rapid and non-destructive solution for road subsurface distress (RSD) detection. However, recognizing RSD from GPR images is labor-intensive and heavily relies on the expertise of inspectors. Deep learning-based automatic RSD recognition, though ameliorating the burden of data processing, suffers from insufficient capability to recognize defects. In this study, a novel cross-verification strategy was proposed to fully exploit the complementary abilities of region proposal networks in object recognition from different views of GPR images. Following this strategy, three YOLO-based models were used to detect the RSD (voids and loose structures) and manholes. Each model was trained with a specific view of 3D GPR dataset, which contains rigorously validated 2134 samples of diverse types obtained through field scanning. The cross-verification strategy achieves outstanding accuracy with a recall of over 98.6% in the tests using real field-scanning data. Field tests also show that deep learning-based automatic RSD recognition can reduce the human labor of inspection by around 90%.
- 6 May 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
A Two-Stage Deep Learning Method with Novel Damage Sensitive Features for Structural Damage Identification
In this paper, a two-stage damage detection method is proposed to integrate unsupervised and supervised approaches, leveraging their complementary strengths to enhance the method’s capability in damage identification. In the first stage, an unsupervised damage detection method based on convolutional autoencoders (CAEs) is introduced to preliminarily determine whether a signal is abnormal. It is trained solely on normal signals and subsequently reconstructs future signals. In addition to the reconstruction error analysis/indicator, a novel damage indicator referred to as the signal-to-reconstruction error ratio (SRER) is introduced to evaluate the reconstruction accuracy and identify potential anomalies. By monitoring the evolution of these two indicators over time, the method achieves 100% accuracy in detecting structural anomalies. In the second stage, to improve efficiency and mitigate the impact of noisy signals on detection results, two damage indicators are used as inputs for supervised deep learning models, achieving over 90% accuracy in damage localization with these two damage-sensitive features. Furthermore, using these two features as inputs improved accuracy by 2–30% compared to conventional vibration inputs. The proposed method and novel damage indicators demonstrate the ability to rapidly detect and accurately locate structural damage, showcasing significant potential for practical engineering applications.
- 6 May 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
A Two-Stage Deep Learning Method with Novel Damage Sensitive Features for Structural Damage Identification
In this paper, a two-stage damage detection method is proposed to integrate unsupervised and supervised approaches, leveraging their complementary strengths to enhance the method’s capability in damage identification. In the first stage, an unsupervised damage detection method based on convolutional autoencoders (CAEs) is introduced to preliminarily determine whether a signal is abnormal. It is trained solely on normal signals and subsequently reconstructs future signals. In addition to the reconstruction error analysis/indicator, a novel damage indicator referred to as the signal-to-reconstruction error ratio (SRER) is introduced to evaluate the reconstruction accuracy and identify potential anomalies. By monitoring the evolution of these two indicators over time, the method achieves 100% accuracy in detecting structural anomalies. In the second stage, to improve efficiency and mitigate the impact of noisy signals on detection results, two damage indicators are used as inputs for supervised deep learning models, achieving over 90% accuracy in damage localization with these two damage-sensitive features. Furthermore, using these two features as inputs improved accuracy by 2–30% compared to conventional vibration inputs. The proposed method and novel damage indicators demonstrate the ability to rapidly detect and accurately locate structural damage, showcasing significant potential for practical engineering applications.
- 29 February 2024
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
- Vol. 22 (03)
Artificial Intelligence-Based Damage Identification Method Using Principal Component Analysis with Spatial and Multi-Scale Temporal Windows
Previous studies have demonstrated the superior damage identification performance of the double-window principal component analysis (DWPCA) method over traditional PCA methods and other traditional techniques, such as wavelet and regression analysis. DWPCA uses temporal windows to discriminate structural states and spatial windows to exclude damage-insensitive responses, making it more effective for damage identification. However, determining the optimal temporal window scale and its impact on damage identification performance still remains unclear. In this study, different scales of temporal windows, including yearly, seasonal and monthly windows, are employed to obtain corresponding damage features, i.e., eigenvectors derived from DWPCA. These damage-sensitive eigenvectors from various temporal windows are then used as inputs for artificial intelligence (AI) algorithms to localize and quantify damages. In this paper two types of AI algorithms are employed: random forest (RF) and bidirectional gated recurrent unit (BiGRU). A numerical study using a benchmark model is used to evaluate the contribution of the eigenvector of each temporal scale to damage identification. The results demonstrate that the combined DWPCA eigenvectors [ ] from the three temporal windows effectively enhance the AI-based damage identification capability. Besides, AI algorithm with [ ] can have high accuracy exceeding 95% under limited training data sets and strong noise. Additionally, when DWPCA eigenvectors from monthly or seasonal windows as inputs, which is both sensitive to damages and noise, the BiGRU also achieves high accuracy of over 90% for damage identification, due to its advantages in feature extraction. These findings suggest that the proposed approach has significant potential for real-life structural health monitoring applications.
- 29 February 2024
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
- Vol. 22 (03)
Artificial Intelligence-Based Damage Identification Method Using Principal Component Analysis with Spatial and Multi-Scale Temporal Windows
Previous studies have demonstrated the superior damage identification performance of the double-window principal component analysis (DWPCA) method over traditional PCA methods and other traditional techniques, such as wavelet and regression analysis. DWPCA uses temporal windows to discriminate structural states and spatial windows to exclude damage-insensitive responses, making it more effective for damage identification. However, determining the optimal temporal window scale and its impact on damage identification performance still remains unclear. In this study, different scales of temporal windows, including yearly, seasonal and monthly windows, are employed to obtain corresponding damage features, i.e., eigenvectors derived from DWPCA. These damage-sensitive eigenvectors from various temporal windows are then used as inputs for artificial intelligence (AI) algorithms to localize and quantify damages. In this paper two types of AI algorithms are employed: random forest (RF) and bidirectional gated recurrent unit (BiGRU). A numerical study using a benchmark model is used to evaluate the contribution of the eigenvector of each temporal scale to damage identification. The results demonstrate that the combined DWPCA eigenvectors [ ] from the three temporal windows effectively enhance the AI-based damage identification capability. Besides, AI algorithm with [ ] can have high accuracy exceeding 95% under limited training data sets and strong noise. Additionally, when DWPCA eigenvectors from monthly or seasonal windows as inputs, which is both sensitive to damages and noise, the BiGRU also achieves high accuracy of over 90% for damage identification, due to its advantages in feature extraction. These findings suggest that the proposed approach has significant potential for real-life structural health monitoring applications.