Computer-Vision and Machine-Learning-Based Seismic Damage Assessment of Reinforced Concrete Structures

Xu, Yang; Li, Yang; Zheng, Xiaohang; Zheng, Xiaodong; Zhang, Qiangqiang

doi:10.3390/buildings13051258

Cited by 19 publications

(5 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the quantification of the seismic damage index, images and experimental data of 124 RC columns during the entire quasi-static experiment process are utilized [19]. The results show that the established regression model of the seismic damage index is unbiased and stable without overfitting.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Coarse-to-Fine Seismic Assessment Based on Computer Vision

XU,

ZHANG,

2023

Proceedings of the 14th International Workshop on Structural Health Monitoring

View full text Add to dashboard Cite

Recently, remote sensing satellites, unmanned aerial vehicles (UAVs), and smartphones have been extensively utilized in non-contact post-earthquake inspection at different scales with cutting-edge computer vision and machine learning techniques. This study establishes a computer-vision-based coarse-to-fine seismic assessment framework to localize dense buildings in urban areas, classify collapsed and noncollapsed states, recognize multi-type surface damage on structural components, and evaluate seismic performances. First, a Transformer-CNN deep learning architecture is designed for semantic segmentation of dense buildings and binary classification of collapsed states using large-scale remote sensing satellite images. It consists of a Swin Transformer encoder, multi-stage feature fusion module, and UPerNet decoder to extract global correlations and local features of dense buildings synchronously. Then, a multi-task learning strategy is proposed to simultaneously recognize multi-type structural components (column, beam, wall), seismic damage (concrete crack, spalling, and rebar exposure), and multi-level damage states (minor, moderate, major) using medium-scale UAV images. It contains a CNN-based encoder-decoder backbone with skip-connection modules and multi-head segmentation subnetworks for different tasks. The geometric consistency loss of split line, area, and curvature is further designed to refine the semantic segmentation of local details, increase boundary smoothness, and suppress inner voids. Finally, a machine learning neural network is established to quantify the seismic damage index of structural components using damage-related parameters (lengths, areas, and numbers of concrete crack, spalling, and rebar exposure) and design-related parameters (axial compression ratio, shear span ratio, and volumetric stirrup ratio) as inputs. A seismic damage indicator with an explicit bound of [0,1] can be obtained to reflect the nonlinear accumulation of seismic damage. The effectiveness and applicability under real-world post-earthquake scenarios have been validated by the 2017 Mexico City Earthquake M7.1, 2008 Beichuan Earthquake M8.0, 2010 Yushu Earthquake M6.9, 2015 Nepal Earthquake M7.8, 2016 Ecuador Earthquake M7.8, and 2016 Meinong Earthquake M6.7.

show abstract

Section: Resultsmentioning

confidence: 99%

“…[18,19]. First, a synthetical indicator of seismic damage with an explicit bound of [0,1] is designed based on refined Park-Ang model.…”

mentioning

confidence: 99%

Coarse-to-Fine Seismic Assessment Based on Computer Vision

XU,

ZHANG,

2023

Proceedings of the 14th International Workshop on Structural Health Monitoring

View full text Add to dashboard Cite

show abstract

“…Taking the S2 stay cable as an example, the variation in the tensile force can be represented by Formula (5).…”

Section: The Application Of Grey Relational Analysis In Determining T...mentioning

confidence: 99%

“…Zain et al [3,4] proposed a novel framework for assessing the vulnerability of tubular structures using machine learning algorithms and investigated the effectiveness of using vulnerability information for tubular buildings through machine learning. Xu et al [5] assessed the seismic damage to reinforced concrete structures using computer vision and machine learning. Asgarkhani et al [6] applied machine learning to the prediction of residual drift and seismic risk in moment-resisting frames considering soil-structure interactions.…”

Section: Introductionmentioning

confidence: 99%

A Machine Learning Model to Predict the Seismic Lifecycle Behavior of a Cross-Sea Cable-Stayed Bridge

Lu,

Liu,

Zhang

2024

Buildings

View full text Add to dashboard Cite

Cross-sea cable-stayed bridges encounter challenges associated with cable corrosion and cable-force relaxation during their service life, which significantly affects their structural performance and seismic response. This study focuses on a cross-sea cable-stayed bridge located in Hainan Province. Utilizing an LSTM deep learning model, this study aims to fill in the gaps in short-term cable-monitoring data from the past year using the available cable-force-monitoring data from the same period. The authors of this study interpolated the cable-force data in the absence of sensors and employed a SARIMA machine learning time-series-prediction model to predict the future trends of all cable forces. A finite-element model was constructed, and a dynamic time-history analysis of the seismic response of the cross-sea cable-stayed bridge was conducted, considering the influence of cable-force relaxation and cable corrosion in the future. The findings indicate that the LSTM-SARIMA model predicted an average decrease of 11.81% in the cable force of the cable-stayed bridge after 20 years. During the lifecycle of the cables, cable corrosion exerts a significant impact on the variation in cable stress within the bridge structure during earthquakes, while cable-force relaxation has a more pronounced effect on the vertical displacement of the main beam of the bridge structure during seismic events. Compared to when using the traditional model that only considers cable corrosion, the maximum negative vertical displacement of the main beam increases by 29.7% when using the proposed model if the earthquake intensity is 0.35 g after 20 years, which indicates that the proposed machine learning model can exactly determine the seismic behavior of the lifecycle cross-sea cable-stayed bridge, considering the impacts of both cable-force relaxation and cable corrosion.

show abstract

“…As a result, artificial and convolutional neural networks, as well as support vectors, proved to be highly efficient and accurate. No detailed analysis of each work is needed to make the conclusion that loads and components of the stress-strain state are identified with the help of artificial intelligence in works on seismic risk analysis [17][18][19][20] and in works on the evaluation of building operation modes and information modeling [21][22][23][24].…”

Section: Introductionmentioning

confidence: 99%

Load Identification in Steel Structural Systems Using Machine Learning Elements: Uniform Length Loads and Point Forces

Tusnin,

Alekseytsev,

Tusnina

2024

Buildings

View full text Add to dashboard Cite

Actual load identification is a most important task solved in the course of (1) engineering inspections of steel structures, (2) the design of systems rising or restoring the bearing capacity of damaged structural frames, and (3) structural health monitoring. Actual load values are used to determine the stress–strain state (SSS) of a structure and accomplish various engineering objectives. Load identification can involve some uncertainty and require soft computing techniques. Towards this end, the article presents an integrated method combining basic provisions of structural mechanics, machine learning, and artificial neural networks. This method involves decomposing structures into primitives, using machine learning data to make projections, and assembling structures to make final projections for steel frame structures subjected to elastic strain. Final projections serve to identify parameters of point forces and loads distributed along the length of rods. The process of identification means checking the difference between (1) weight coefficient matrices applied to unit loads and (2) actual loads standardized using maximum load values. Cases of neural network training and parameters identification are provided for simple beams. The aim of this research is to enhance the reliability and durability of steel structures by predicting consequences of unfavorable load, including emergency impacts. The novelty of this study lies in the co-use of artificial intelligence elements and structural mechanics methods to predict load parameters using actual displacement curves of structures. This novel approach will enable engineering inspection teams to predict unfavorable load peaks, prevent emergency situations, and identify actual causes of emergencies triggered by excessive loading.

show abstract

Computer-Vision and Machine-Learning-Based Seismic Damage Assessment of Reinforced Concrete Structures

Cited by 19 publications

References 60 publications

Coarse-to-Fine Seismic Assessment Based on Computer Vision

Coarse-to-Fine Seismic Assessment Based on Computer Vision

A Machine Learning Model to Predict the Seismic Lifecycle Behavior of a Cross-Sea Cable-Stayed Bridge

Load Identification in Steel Structural Systems Using Machine Learning Elements: Uniform Length Loads and Point Forces

Contact Info

Product

Resources

About