Operation-aided system for fault diagnosis of continuous and semi-continuous processes

Knowledge-based fault diagnostic systems for aromatic process and polypropylene process, named FINDS/AC and FINDS/PP for short, are presented together with their application examples. Knowledge bases for diagnosis are derived from the signed digraph. In the FINDS/AC, the process is decomposed into several subsystems and corresponding conduit systems to achieve the easiness of knowledge base maintenance and the adaptiveness for real-time environment. Through off-line analysis of the process digraph, sympton-fault associations are obtained mainly for diagnostic speed, and more suitable fault propagation paths are determined using the information such as the length of propagation time and the strength of causal interactions. In the FINDS/PP, two types of model are used for diagnosis. The RCED (Reduced Cause Effect Digraph) based on the signed digraph is constructed and restored in knowledge base. The PGTT (Pattern Graph Through Time) is generated on the real-time basis during the diagnosis period. The PGTT represents cause-effect relationships for time among measured state variables and is effectively able to treat nonsingular transition. It allows diagnosis for unsteady-state under limited conditions. This algorithm is applied to the diagnosis system for polypropylene process.