Monitoring of multi-bolt connection looseness using a novel vibro-acoustic method

Bolted connections are prone to losing their preloads with the increasing service life, thus inducing engineering accidents and economic losses in industries. Therefore, it is important to detect bolt loosening, while current structural health monitoring methods mainly focus on single-bolt joints, whose applications in industries are limited. Thus, in this paper, a novel vibro-acoustic modulation (VAM) method, is developed to detect looseness of the multi-bolt connection. Compared to traditional VAM, the proposed method uses linear swept sine waves for both low-frequency and high-frequency excitations, which avoids a priori knowledge of the structure. Moreover, the orthogonal matching pursuit method is applied to compress original modulated signals and exclude redundant features. Then, a new entropy, namely the Gnome entropy with acronym gEn , is proposed in this paper. According to simulation analysis, the gEn has better anti-noise capacity and fewer parameters than traditional entropy. Finally, after quantifying the dynamic characteristics of compressed signals to obtain feature sets through the gEn , we feed feature sets into a random forest classifier and achieve looseness detection of the multi-bolt connection. Moreover, the proposed method in this paper has great potential to detect other structural damages and provides guidance for further investigations on the VAM method.