Real-time structural health monitoring of steel structures using acoustic emission signals and a KAN-LSTM deep learning framework

Jialin Cui, Chunwang Lv, Jinbo Du,
Real-time structural health monitoring of steel structures using acoustic emission signals and a KAN-LSTM deep learning framework,
Engineering Structures,
Volume 344,
2025,
121328,
ISSN 0141-0296,
https://doi.org/10.1016/j.engstruct.2025.121328.
(https://www.sciencedirect.com/science/article/pii/S0141029625017195)
Abstract: This study presents an advanced deep learning framework for real-time structural health monitoring (SHM), addressing the urgent need for accurate fatigue damage classification in steel structures under dynamic service conditions. The proposed framework integrates Kolmogorov-Arnold Networks (KAN) with Long Short-Term Memory (LSTM) networks to process Acoustic Emission (AE) signals captured from a steel structure under tensile stress. AE signals, recorded through piezoelectric sensors, were categorized into three distinct damage stages: crack initiation (CI), crack growth (CG), and crack fracture (CF). Given the sequential nature of AE data, signals were divided into fixed-length segments, which were subsequently processed using the LSTM module to capture both short- and long-term dependencies. The experimental results demonstrate that the KAN-LSTM framework outperforms traditional models, including KAN, KAN-Recurrent Neural Network (RNN), and KAN-Gated Recurrent Unit (GRU), in terms of classification accuracy. Specifically, the KAN-LSTM model achieved a peak classification accuracy of 98.7 %. Sensitivity analysis revealed that optimal performance was obtained when the signal segment length was set to 8, yielding classification accuracies of 97.9 %, 100 %, and 100 % for CI, CG, and CF states, respectively, with an overall accuracy of 99.3 %. These findings demonstrate the framework’s effectiveness in capturing key temporal features for accurate real-time damage detection, supporting its practical use in SHM.
Keywords: Structural Health Monitoring; Acoustic Emission; Deep Learning; Kolmogorov-Arnold Network; Long Short-Term Memory; Damage Classification