A Methodology for Knowledge Acquisition and Reasoning About Faults in Conditions of Incomplete Information

Abstract

In this paper the problem of modelling knowledge about fault propagation and utilizing this model for reasoning in conditions of incomplete information is addressed. The term incomplete information is used to indicate the situations when available amount of information isn’t enough to build well known analytical or statistical models for problem of failure analysis. A solution that merges digraph and event tree failure analysis with three types of reasoning is presented. The methodology is divided into two phases. The first phase is structural modelling phase and the second phase is the model utilization phase for cause-consequence rule derivation. The results of these phases are the topological knowledge base and the deep knowledge rule base, respectively. The two knowledge bases are combined to form a novel hierarchical knowledge representation scheme. Upper levels of this hierarchy are made up of the structural models visualized by digraphs. Each level represents the system from the point of view of morphological and functional structure, respectively. The lowest level consists of a set of event trees. The two knowledge bases support three kinds of reasoning, namely, reasoning in logic, diagnostic and predictive reasoning. Formal methods for model building, event tree construction and derivation of cause-consequence rules are discussed.