GENERATION OF DIAGNOSTIC TREES BY MEANS OF SIMPLIFIED PROCESS MODELS AND MACHINE LEARNING

Fault diagnosis by means of diagnostic trees is of considerable intere st for industrial applications. The drawbacks of this approach are mos tly related to the knowledge elicitation through laborious enumeration of the tree structure and ad hoc threshold selection for symptoms def inition. These problems can be alleviated if a more profound knowledge of the process is brought into play. The main idea of the paper consi sts of modeling the nominal and faulty states of the plant by means of interval-like component models derived from first-principles laws, e. g. the conservation law. Such a model serves to simulate the entire sy stem under different fault conditions, in order to obtain the represen tative patterns of measurable process quantities, i.e. training exampl es. To march these patterns by diagnostic rules, multistrategy machine learning is applied. As a result, binary decision trees that relate s ymptoms to faults are obtained, along with the thresholds defining the symptoms, This technique is applied to a laboratory test process oper ating in the steady state, and is shown to be suitable for handling in cipient single faults. The proposed learning approach is compared with two related machine learning methods, It is found that it achieves si milar classification accuracy with better transparency of the resultin g diagnostic system. (C) 1997 Elsevier Science Ltd. All rights reserve d.