Detection and Localization of Looseness in Multiple Bolts Using Time Reversal Signal and Neural Network Techniques

Bolt connected joints are commonly used in many industries. Detection the connection conditions of bolts, especially in the early damage stage, is essential for ensuring the safety and stability of the whole structure. A shear horizontal (SH) guided wave-based bolt looseness detection method, excited by a small number of magnetostrictive patch transducer, is presented in this paper by combining the time reversal signal and the back propagation neural network techniques. The damage index was proposed based on the peak amplitude of the reconstructed response excited by the time reversal signal of the measured unit impulse response. The damage severities and damage locations can be well estimated using the damage indices from several wave propagation paths. Numerical simulations were carried out to generate the training patterns with various damage severities and train the BPNN. Experimental tests were conducted to test the accuracy of the trained BPNN. The local damage severity can be well estimated using the BPNN trained by the training patterns generated from the finite element model.