Deep learning-based intelligent evaluation of freeze-thaw damage in tunnel lining using PZT-induced stress wave

Abstract

Tunnel lining structures in cold regions are frequently damaged by freeze-thaw (f-t) cycles, which have emerged as one of the main factors to affect the service life of tunnels. Real-time and accurate evaluation of such damage is of great importance for the development of effective maintenance strategies. In this paper, we proposed a novel deep learning-based intelligent method to evaluate the f-t damage in tunnel lining using PZT-induced stress wave. First of all, f-t cycle durability experiments were conducted on the tunnel lining concrete, during which the stress wave signals were measured using PZT transducers. The waveform and energy change rule of the stress wave signals during the f-t cycle were analyzed. After that, the time-frequency diagrams were generated from the collected signals through the continuous wavelet transform (CWT). Meanwhile, a fine-scale numerical model was proposed to simulate concrete f-t damage process, and the damage stages were divided based on the simulated damage unit area. As a result, a CWT-based image dataset with damage stage labels was established. In addition, a convolutional neural network (CNN) model was developed for automatically extracting features from the CWT images and ultimately predicting the f-t damage stage. Finally, the performance of the CNN model was compared with two machine learning algorithms including support vector machine (SVM) and back-propagation neural network (BPNN). The results show that the CNN model can accurately predict the concrete f-t damage stage and has superior performances over other methods, indicating its great potential for real-time monitoring and rapid evaluation of concrete f-t damage in tunnel lining.