Abstract
In response to the difficulties faced in detecting bolt connection damage in steel truss structures, this paper proposes a bolt loosening identification method based on sound signal analysis, a Genetic Algorithm-Optimized Support Vector Machine (GA-SVM), and Recursive Feature Elimination (RFE). By preprocessing and feature extraction of sound signals, short-term energy, short-term zero crossing rate, and wavelet packet frequency band energy features were extracted. SVM-RFE was used for sensitive feature selection, and genetic algorithm was combined to optimize SVM parameters, ultimately obtaining the optimal recognition model. The effectiveness of this method was verified through bolt loosening tests on steel truss structures. The results showed that the method can achieve a recognition accuracy of 99.5% with a small training dataset, and has strong practicality and feasibility, providing technical support for safety monitoring of engineering structures.