An expert system for fault diagnosis in transformers during impulse tests

Abstract

The objective of this paper is to present an expert system for fault diagnosis in transformers during impulse tests. Transformers are impulse-tested by applying a specific set of test impulse sequences and the resulting voltage and current oscillograms are recorded. The presence of any fault in the transformer winding is detected using these recorded oscillograms. These complex decision tasks are, in general, performed by experienced testing personnel, whose knowledge can be expressed as a set of production rules. The knowledge base of the expert system includes as many as possible of these production rules. To identify, characterize and locate a fault, an inference engine is developed to perform deductive reasoning based on the rules in the knowledge base and the statistical techniques. The basic idea of the expert system is to make a fault diagnosis in a user-friendly windowed environment. The modularity of the system enables easy expansion and modification of the database, the rule-base as well as the inference engine.