Diagnosis for Fatigue Cracking in Concealment of Large-Scale Overloaded Supporting Shaft Based on Time Series and Fuzzy Clustering

Fatigue cracking in concealment of large-scale overloaded supporting shaft often causes accident, breaks product line, and results in high loss for enterprises due to the unobserved and random characteristics. A novel approach is proposed to detect the fatigue cracking based on time series and fuzzy clustering. The vibration signal with structural state information can be collected from its pedestal position through analyzing the features of such fatigue cracking. The AR diagnosis model is established by using the time series method based on MATLAB procedure. The model parameters with integrated failure information are taken as feature values to form standard failure mode feature vectors space. The fuzzy clustering theory is applied to achieve the failure pattern recognition, and combined with the Euclidean distance discriminant function to match the candidate vector with maximal subjection degree among the standard vector space. Thus, the fatigue cracking of measuring supporting shaft can be diagnosed. In the end, the result of an example demonstrates that the proposed approach is effective and easy to implement.