A rule‐based approach to fault diagnosis using the signed directed graph

Abstract

Abstract Fault diagnosis is the problem of determining the root causes of process upsets. This paper presents a very efficient method of identifying the possible causes of process disturbances using the signed directed graph (digraph) representation of process interactions. The analysis is based on forming logical statements (rules) derived from the process digraph; these are evaluated using on‐line data to yield the diagnosis. Evaluation of rule antecedents is more efficient than the previous algorithmic approach of Shiozaki et al. In the rule‐based approach, the diagnostic criteria are represented explicitly, not hidden by a complex algorithmic procedure. This allows the diagnostic rules to be tailored to reflect the best available knowledge of plant behavior. The rules generated by this technique can be integrated with other rules on plant operations using an expert systems framework.