Underground All-Dielectric Self-Supporting (ADSS) optical cables in urban areas are frequently compromised by external disturbances such as construction activities and human excavation, posing a serious threat to power grid security. Current single-algorithm approaches often struggle to adapt to the complex and variable urban environment, resulting in limited recognition accuracy. To address these challenges, this paper introduces LSTM-CNN-CatBoost-GSSSA, a hybrid recognition model that effectively captures the temporal, frequency, and spatial propagation characteristics of vibration signals. It integrates the sequential modeling capability of Long Short-Term Memory (LSTM) networks with the local feature extraction power of Convolutional Neural Networks (CNN). This layer generates adaptive weights through differentiable learning to realize the dynamic weighted fusion of time series and spatial-frequency domain features, thereby optimizing the feature selection mechanism. Moreover, an improved golden sine-enhanced Sparrow Search Algorithm (GSSSA) is employed to globally optimize CatBoost's hyperparameters and adaptively adjust feature weights for enhanced recognition. Experimental results demonstrate that the proposed model achieves an accuracy of 97.6%, surpassing SVM, CNN, LSTM-CatBoost, CNN-CatBoost, LSTM-CNN-GBDT, and LSTM-CNN-CatBoost by 15%, 3.6%, 14.79%, 1.2%, 1.4%, and 0.49%, respectively. The model exhibits both high recognition performance and good computational efficiency, providing a feasible technical solution for real-time monitoring and intelligent diagnosis of external force damage to urban buried ADSS optical fiber cables. This advancement contributes to improving the operational safety and reliability of power communication networks.
Chen et al. (Mon,) studied this question.