Dual-chanxnel deep learning based cable tunnel damage identification and localization technique

Purpose Cable tunnels are crucial infrastructures in urban areas, supporting power distribution, communication and transportation systems. Ensuring their structural health is vital for maintaining safety and reliability. Vibration signals are widely used for monitoring structural conditions; however, anomaly detection and damage localization in such complex environments remain challenging due to noise interference and the non-stationary nature of the signals. Design/methodology/approach This paper presents a deep learning framework for anomaly detection and damage localization in cable tunnels using vibration signals. The authors propose an adaptive CEEMDAN-wavelet denoising (ACWD) algorithm to effectively remove noise and improve signal quality. In addition, a partial-whole double sliding window (PWDSW) algorithm is introduced for disturbance preclassification, reducing false positives and enhancing model efficiency. The core model, CNN–GRU–CBAM, combines convolutional neural networks (CNN), gated recurrent units (GRU) and the convolutional block attention module (CBAM) attention mechanism to capture both spatial and temporal features of the signals. Findings Experimental results show that the proposed model outperforms existing methods in terms of accuracy, recall and MAE. The framework provides a robust and scalable solution for real-time cable tunnel health monitoring, improving both anomaly detection and damage localization. Originality/value Vibration signals have long been used to monitor cable tunnels, but their effectiveness has been limited by noise and misclassification. The authors show that a novel signal processing pipeline, combining ACWD with a PWDSW preclassifier, can significantly improve anomaly detection accuracy. To the best of the authors’ knowledge, for the first time, the authors demonstrate how spatial-temporal features extracted by a CNN–GRU–CBAM network can localize damage while filtering out transient disturbances.