Abstract:
Based on the broad-band random vibration experiment on a structure model of clamping support used in transportation, this paper investigates the problems on detecting the time, evaluating the extent and locating the place of the attachment bolt looseness in the structure model. Firstly, wavelet transform was used to analyse the structural acceleration response data, which includes the moment when the bolt looseness occurred. An obvious step leap along the time axis, which shows the time when the bolt looseness occurs, can be clearly observed in the level-1 details of the wavelet decomposition. Secondly, ARX (auto-regression with exogenous inputs) model was constructed from the measured input and output signals, then the statistical indices of the model residual error were defined as the damage characteristic parameter indicating the severity and location of the loose bolts in the clamping support. The experimental study shows that the extracted characteristic parameter in this paper can be more appropriate than measuring the base frequency for monitoring the bolts looseness at early stage.
