Employing Acoustic Emission for Cost-Effective Continuous Monitoring of a Multi-Bolt Joint

Authors

1 University of Gonabad

2 university of ganabad

10.22044/jsfm.2026.16503.3992

Abstract

Bolt loosening in mechanical joints remains one of the critical challenges in maintaining the structural integrity of industrial systems. In this study, a novel approach is proposed for monitoring the loosening status of bolts in a multi-bolt joint using low-sampling-rate acoustic emission (AE) signals. An experimental setup consisting of a four-bolt connection was designed, and acoustic signals were recorded under sixteen different bolt-tightening configurations. To analyze the signals, Mel-frequency cepstral coefficients (MFCCs) were extracted as feature vectors, and the root mean square deviation (RMSD) index was employed to quantify signal variations. The results showed that bolt loosening led to a noticeable increase in RMSD compared to the healthy state. However, no significant correlation was observed between the number of loosened bolts and RMSD values. Subsequently, five classification scenarios were designed, and the performance of a feedforward neural network was evaluated. The highest classification accuracy of 94.44% was achieved in the scenario where connections with one loosened bolts or more were separated from the rest. The proposed method, while relying on simple hardware and lightweight data, demonstrated high accuracy in the early detection of bolt loosening and shows strong potential for integration into continuous structural health monitoring systems in industrial environments.

Keywords

Main Subjects