Bolt looseness detection and localization using wave energy transmission ratios and neural network technique

Looseness detection in bolt-connected joints is vital in ensuring safety and keeping the service stability of structures. Thus, various structural health monitoring methods have been introduced for bolt looseness detection by many researchers. However, most of them studied a single bolt, which may not be readily applicable to actual structures. In this study, a SH-type guided wave-based method is presented for bolt looseness detection and localization of a joint with multiple bolts using a small number of magnetostrictive transducers. A normalized wave energy transmission ratio IBLnor was used as a bolt looseness index, which was defined on the basis of the wave energy ratios between the transmitted wave passing through the joint and the directly incoming wave from the actuator. Several wave propagation paths in the pitch-catch tests were considered, and the IBLnor values from the wave paths were used as the input to the backpropagation neural network (BPNN) for bolt looseness localization and severity estimation. Numerical and experimental studies were conducted on a lap joint with eight bolts. The results show that the bolt looseness conditions can be successfully estimated for the experimental data using the BPNN trained by the IBLnor generated from the finite element simulation. Noise-injected learning was conducted in the training process to improve the bolt looseness localization accuracy.