Application of principal component analysis for fault detection of DC motor parameters

Abstract

The aim of technical processes supervision is to reveal the actual system state and also undesired states to be indicated. The deviations from normal process behavior are usually results of faults. They might cause malfunctions or failures that must be prevented. The monitoring of a technical process in normal operation state is a part of the supervision process. This is usually performed by a limit-checking of some measurable output variables. Certain alarms rise if the variables values are found outside of the tolerance zone set. The main purpose of the principal component analysis (PCA) is to reduce the dimensionality of a data set containing a large number of interrelated variables, retaining at the same time as much as possible of the variations presented in the initial data set. This reduction is achieved by transformation to a new set of the uncorrelated variables called principal components. They are arranged so that the first few of them retain the most of the variations of all of the original variables presented. This paper presents an application of the principal component analysis for real time fault detection of DC motor parameters.