Fault detection using topological case based modelimg and its application to chiller performance deterioration

Abstract

We apply a new modeling technique, Topological Case Based Modeling (TCBM), to fault diagnosis. In this paper we propose a new model which represents a continuous input/output relation using a set of numerical data, and it is possible to describe nonlinear systems. We employ the idea of case based reasoning (CBR). CBR infers a new case from stored cases and these relation.[1] It is important to define the similarity, that is a relation among the cases. We propose to define the similarity as the neighbourhood in input space corresponding to output accuracy and its accuracy is arbitrarily set before constructing the model. We name this technique Topological Case Based Modeling. In addition, we describe that TCBM has several advantages over other models. Finally we also show the example of TCBM to detect the deterioration for a chiller system.<>