Geometric Attention Regularization Enhancing Convolutional Neural Networks for Bridge Rubber Bearing Damage Assessment

Cui, Mida; Wu, Gang; Chen, Zhi Qiang; Dang, Ji; Zhou, Mei; Feng, Dongming

doi:10.1061/(asce)cf.1943-5509.0001634

Cited by 9 publications

(3 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, a convolution-based multi-damage recognition neural network combined CNN with an attention network and hybrid pooling layers was developed by Shin et al (2020) to classify the five damage types and an accuracy of 98.9% was achieved. Cui et al (2021a) proposed a geometric attention regulation method, in which the bearing location information was marked by a bounding box worked as an attention mechanism to indicate the important part of the input image. The experiments proved that the method could enhance CNN's performance effectively.…”

Section: Image Classificationmentioning

confidence: 99%

The application of deep learning in bridge health monitoring: a literature review

Zhang

Wang

et al. 2022

ABEN

View full text Add to dashboard Cite

Along with the advancement in sensing and communication technologies, the explosion in the measurement data collected by structural health monitoring (SHM) systems installed in bridges brings both opportunities and challenges to the engineering community for the SHM of bridges. Deep learning (DL), based on deep neural networks and equipped with high-end computer resources, provides a promising way of using big measurement data to address the problem and has made remarkable successes in recent years. This paper focuses on the review of the recent application of DL in SHM, particularly damage detection, and provides readers with an overall understanding of the missions faced by the SHM of the bridges. The general studies of DL in vibration-based SHM and vision-based SHM are respectively reviewed first. The applications of DL to some real bridges are then commented. A summary of limitations and prospects in the DL application for bridge health monitoring is finally given.

show abstract

Section: Image Classificationmentioning

confidence: 99%

The application of deep learning in bridge health monitoring: a literature review

Zhang

Wang

et al. 2022

ABEN

View full text Add to dashboard Cite

show abstract

“…Peel et al (2018) presented a modified DiddyBorg robotic to bridge bearing inspection combining visual sensors and Hector SLAM. Cui et al (2021) developed a geometric attention regularization method to enhance the performance of CNN for the condition evaluation of rubber bearings. Su et al (2023) developed a real-time simultaneous measurement method of bridge bearing displacement and rotation based on feature-constrained visual odometry to solve the bottleneck problem of existing displacement and rotation measurement methods.…”

Section: Introductionmentioning

confidence: 99%

Bearing condition assessment of a simply elastic supported beam based on impact vibration testing

Xia,

Zhou,

Yang

et al. 2024

Advances in Structural Engineering

View full text Add to dashboard Cite

The mechanical properties of bridge bearings gradually deteriorate over time, resulting from daily traffic loading and harsh environmental conditions. However, efficient detection of in-service bridge bearings is still challenging, especially in quantifying support stiffness. This study presents a bridge bearing condition assessment framework of a simply elastic supported beam based on impact vibration testing and structural flexibility identification. Firstly, the dynamic response equation of a simply supported beam with elastic boundary constraints is established. The modal orthogonality of Euler-Bernoulli beams and the corresponding overall matrix of undetermined coefficients are derived using the separate variable method. Subsequently, the flexibility matrix of the simply elastic supported beam is derived and verified in the cases of forward and inverse analysis, respectively. Furthermore, the vertical and rotational stiffness of the bearings are established based on the structural flexibility matrix. Finally, experimental verification is performed to identify the flexibility and the vertical support stiffness of the simply elastic supported beam based on impact vibration tests. The results indicate that the proposed the bearing condition assessment framework can straightforwardly quantify support stiffness.

show abstract

“…It can reduce the attention of the model to useless information and emphasize the role of important features through weight allocation. And many studies have proved that the attention mechanism can improve the prediction accuracy and reliability of time series modelling [23][24][25] . Therefore, an attention mechanism-based long short-term memory neural network (AttLSTM) model is applied to SCB prediction in this study.…”

Section: Introductionmentioning

confidence: 99%

Research on modeling and predicting of BDS3 satellite clock bias using the attention mechanism-based LSTM (AttLSTM) neural network model

Li,

He,

Eken

et al. 2023

JGPS

View full text Add to dashboard Cite

In the Global Navigation Satellite System (GNSS), the satellite clock bias (SCB) plays an important role in the application of real-time precise point positioning (RT-PPP). Based on the operation of Beidou satellite global service, it is very important to establish a reliable Beidou SCB prediction model. In this research, an attention mechanism-based long short-term memory neural network (AttLSTM) model is applied to SCB prediction. The attention mechanism introduced in modelling can make the model pay less attention to useless information through weight allocation. In this paper, the BeiDou-3 Navigation Satellite System (BDS-3) satellite precision clock product provided by GFZ is used for clock prediction experiments. The proposed AttLSTM model, long short-term memory neural network (LSTM) model and quadratic polynomial (QP) model are compared and evaluated, and 12h and 24h SCB prediction experiments of BDS-3 satellite are set up. The results show that AttLSTM model can achieve high SCB prediction accuracy, and the averaged prediction accuracy of 12h and 24h can reach 1.41ns and 1.75ns. Compared with LSTM and QP models, the prediction accuracy of AttLSTM model is improved by 26.1%, 38.4% for 12h and 29.1%, 43.1% for 24h, respectively. Then, the clock bias predicted by the three models is applied to the static PPP positioning experiment, respectively. Through the analysis of the positioning results of 15 MGEX stations, the averaged positioning accuracy of AttLSTM model in the East, North and Up directions can reach 0.074m, 0.019m and 0.154m, respectively. Compared with LSTM and QP models, the positioning accuracy of AttLSTM model is improved by 42.5% and 44.4% in the East direction, 44.7% and 58.9% in the North direction, and 21.7% and 21.8% in the Up direction.

show abstract

Geometric Attention Regularization Enhancing Convolutional Neural Networks for Bridge Rubber Bearing Damage Assessment

Cited by 9 publications

References 29 publications

The application of deep learning in bridge health monitoring: a literature review

The application of deep learning in bridge health monitoring: a literature review

Bearing condition assessment of a simply elastic supported beam based on impact vibration testing

Research on modeling and predicting of BDS3 satellite clock bias using the attention mechanism-based LSTM (AttLSTM) neural network model

Contact Info

Product

Resources

About