Fisher Discriminant Analysis for fault classification

In this paper, a method of fault classification based on Fisher Discriminant Analysis (FDA) for Fault Classification is presented. By using dual FDA to process two sets of data, including normal data and failure data, it is possible to extract discriminative features from overlapping fault data. The method can be applied when the fault data is a bias of a single monitoring parameter. Still, it remains accurate when the fault data is a combination of several parameters deviation. The applicability is discussed by a simulation example. Also, an illustrated application example of this method in the performance data of an aircraft engine is given.