Fault diagnosis method based on robust principal component analysis

Key words:  fault diagnosis  robust principal component analysis  outliers  NIPALS algorithm  M estimator

Abstract:

A robust principal component analysis (PCA) method was proposed to Mialyze the model data with outliers in process monitoring. By replacing the least squares estimator with a robust M-estimator, the traditional principal component analysis problem was transformed into a weighted reconstructed error optimization problem. The problem can be solved by improved nonlinear iterative partial least squares (NIPALS) algorithm so that precise principal component model was available -and monitoring statistics were used to detect faults. The simulation results on a continuous stirred tank reactor ( CSTR) system show that the proposed robust PCA method can remove the influence of outliers, analyze the process data more accurately and diagnose process faults more effectively than traditional PCA method.