Subway tunnel damage detection based on in-service train dynamic response, variational mode decomposition, convolutional neural networks and long short-term memory

Zhang, Yonglai; Xie, Xuansong; Li, Hongqiao; Zhou, Biao; Wang, Qiang; Shahrour, Isam

doi:10.1016/j.autcon.2022.104293

Cited by 32 publications

(9 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in 16, with the increase in the L2 regularization parameters, the accuracies on the training and test sets first increase, peak at a regularization parameter value of 0.0001, and then start to decrease as the regularization parameter continues to increase. This phenomenon demonstrates that it is beneficial to add L2 regularization parameters within a certain range, which The convolution kernel size of (16,32), batch size of 32, learning rate of 0.001, maximum number of iterations of 600, and L2 regularization parameter of 0.0001 are integrated into the previous analysis. The CNN-LSTM structural parameters are presented in table 6.…”

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

“…The configuration results obtained under different schemes of the convolution kernel size of each layer are illustrated in figure 12, where the optional values of the configuration schemes are in four groups, namely group 1 (16,16), group 2 (16,32), group 3 (32,16), and group 4 (32,32). Figure 12 shows that when the abscissa is groups 1 or 2, the accuracies of the training set and test set increase with the increase in the size of the convolution kernel of each layer.…”

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

“…The same phenomenon also occurs in the training time and test time. When the selected convolution kernel size is group 2 (16,32), the accuracies of the training and test sets reach the highest in a short time.…”

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

“…To verify the effectiveness and practicability of the improved NGO-VMD-CNN-LSTM method proposed in this study, the experimental results are compared with those of NGO-VMD-SSA-LSTM, NGO-VMD-WOA-LSTM, VMD-CNN-LSTM, VMD-SSA-LSTM, VMD-WOA-LSTM, CNN-LSTM, SSA-LSTM and WOA-LSTM methods [32][33][34][35]. The optimized parameter settings of each method are presented in table 7.…”

Section: Experimental Comparison and Analysismentioning

confidence: 99%

See 3 more Smart Citations

High accuracy key feature extraction approach for the non-stationary signals measurement based on NGO-VMD noise reduction and CNN-LSTM

Xu,

Jing,

Zhang

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

The effective extraction of key features in non-stationary signals measurement is crucial in numerous engineering fields, including fault diagnosis, geological exploration, and state detection. To accomplish a more accurate and efficient extraction of key feature information from non-stationary signals, we design a novel approach based on variational mode decomposition (VMD) optimization by northern goshawk optimization (NGO) algorithm, convolutional neural network (CNN), and long short-term memory network (LSTM). First, NGO is used to optimize multiple intrinsic mode functions of VMD and reconstruct the signal according to the linear correlation method. Subsequently, the features of moving root mean square, moving kurtosis, and upper envelope are calculated, thereby constructing the feature matrix. Finally, the CNN-LSTM model is established with the chosen optimal hyperparameters prior to the training phase. The experimental results demonstrate that the proposed NGO-VMD-CNN-LSTM method, with a high accuracy reaching 98.22%, can more accurately extract the key information of typical non-stationary signals.

show abstract

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

Section: Discussion Of Cnn-lstm Related Parametersmentioning

confidence: 99%

Section: Experimental Comparison and Analysismentioning

confidence: 99%

See 2 more Smart Citations

High accuracy key feature extraction approach for the non-stationary signals measurement based on NGO-VMD noise reduction and CNN-LSTM

Xu,

Jing,

Zhang

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…During operation of these tunnels, they are exposed to various engineering disturbances (Xie et al, 2020), with the train moving load on the railway being one of the most prominent and impactful disturbances. The dynamic loads generated by trains moving on railways can indeed disturb tunnels beneath them, potentially leading to damage and instability in these tunnels (Yan et al, 2020;Du et al, 2021;Zhang et al, 2022). This poses a signi cant threat to the safety of the tunnels as well as the railways passing above them (Xie et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Experimental study on dynamic response of rock tunnel subjected to train moving load

Zhang

Xie

Guo

et al. 2023

Preprint

View full text Add to dashboard Cite

With the rapid development of transportation engineering, it is increasingly common for railways to closely pass over tunnels. The tunnels might be suffered from damage and destruction under the dynamic loads generated by trains moving on railways, which can in turn threaten the safety of the upper railways. Therefore, it is essential to well understand the dynamic response of tunnels subjected to the upper train moving load. However, there is currently a lack of relevant research on this topic, especially regarding model experiments. To investigate the dynamic response of tunnels subjected to train moving load, physical model experiments were carried out on the train-rail-bridge-shaking table test system, where train moving load was applied through moving of a model train along model tracks. The testing results showed that the train moving speed (vtr) significantly impacts the dynamic response of the tunnel and the surrounding rock mass. Except for the peak particle velocity (PPA) at tunnel vault, the PPAs at other locations and the vibration dominant frequency (f0) at all locations increase as the train moving speed increase. The average PPAs at tunnel spandrels undergo more significant changes compared to the PPAs at tunnel vault, tunnel invert, and inside the surrounding rock mass when the vtr increases. With the exception of some special cases, the average PPAs on tunnel spandrels are generally less than that on the tunnel vault and greater than that on the tunnel invert. As vtr increases, the average f0 on tunnel wall changes more obviously compared to average f0 inside the surrounding rock mass. When the vtr exceeds 10.54 m/s, corresponding to the prototype train moving speed of 120 km/h, the average f0 on tunnel wall are noticeably larger than those inside the surrounding rock. The high coefficients of determination obtained from the nonlinear regression analysis indicates a power relationship between the PPA and vtr, while the f0 shows an exponential association with vtr. The findings of this study hold great importance in ensuring the safe operation of tunnels and railways.

show abstract

Structural damage identification based on variational mode decomposition–Hilbert transform and CNN

Diao

Wang

et al. 2023

J Civil Struct Health Monit

View full text Add to dashboard Cite

Subway tunnel damage detection based on in-service train dynamic response, variational mode decomposition, convolutional neural networks and long short-term memory

Cited by 32 publications

References 37 publications

High accuracy key feature extraction approach for the non-stationary signals measurement based on NGO-VMD noise reduction and CNN-LSTM

High accuracy key feature extraction approach for the non-stationary signals measurement based on NGO-VMD noise reduction and CNN-LSTM

Experimental study on dynamic response of rock tunnel subjected to train moving load

Structural damage identification based on variational mode decomposition–Hilbert transform and CNN

Contact Info

Product

Resources

About