Monitoring of multi-bolt connection looseness using entropy-based active sensing and genetic algorithm-based least square support vector machine

Looseness detection of bolted connections is an essential industrial issue that can reduce the maintenance and repair costs caused by joint failures; however, current loosening detection methods mainly focus on the single-bolt connection. Even though several methods, such as the vibration-based method and electro-mechanical impedance (EMI) method, have been employed to detect multi-bolt looseness, while they are easily affected by environmental issues. Therefore, the main contribution of this paper is to detect loosening of the multi-bolt connection through the PZT-enabled active sensing method, which has several merits including easy-to-implement, low cost, and good ability of anti-environment disturbance. Since the current indicator of the active sensing, namely the signal energy is insensitive to multiple damages, we developed a new damage index (DI) based on the multivariate multiscale fuzzy entropy (MMFE). Subsequently, the maximum relevance minimum redundancy (mRMR) was used to select significant features from the MMFE-based DI to construct the new datasets. After feeding the new datasets into the genetic algorithm-based least square support vector machine (GA-based LSSVM), we trained a classifier to detect loosening of the multi-bolt connection. Finally, repeated experiments were conducted to demonstrate the effectiveness of the proposed method, which can guide future investigations on bolt looseness detection. Published by Elsevier Ltd.