Prediction of bolt fastening state using structural vibration signals

Abstract

We have proposed a new method to predict the state of bolt fastening connection using time-domain structural vibration signal and experimentally validated its effectivness. To obtain the structural vibration signal, non-contact type laser displcement sensor and contact type piezo film sensor were used, respectively. Two-beam structures with holes were prepared and fastened with a set of bolt and nut. By applying a random initial displacement at the free end of the cantilever beam structure, vibration signals were measured for three different bolt fastening states: fully fastened, half-loosened and 90 %-loosened. After extraction of features from the obtained vibration signals, the bolt fastening state was classified based on the k-nearest neighbor (k-NN) algorithm. It is experimentally verified that the bolt fastening state can be accurately predicted by using the structural vibration signals and machine learning algorithm.