Fault Detection and Classification Based on PCA and N-LMC

This paper analyzes the fault detection principle based on principal component analysis (PCA), which uses the upper control limit (UCL) of T² statistic and the quality (Q) control limit of squared prediction error (SPE) statistic to determine whether there is a fault. And then it determines where is wrong by calculating the contributions of each variable to SPE statistic. However, only this cannot identify the type of fault. In connection with this deficiency of the method, this paper proposes a new local mean-based nonparametric classification method (N-LMC) for identifying the fault type after finding the fault data that are detected by fault detection, then the fault detected by PCA can be solved. Simulations on fault detection on the basis of PCA, the N-LMC and the classification of fault data based on PCA and the proposed N-LMC have been carried out in this paper. The simulation results show that classification accuracy of the proposed N-LMC is higher than the existing local mean-based nonparametric classifier (LMC). In addition, the experiments present that combining PCA with the proposed N-LMC can identify the type of fault data more effectively.