Model-based diagnostic expert system for chemical plants

A domain-independent model-based diagnostic system (PRODES) is described. It demonstrates the fact that partitioning the domain knowledge into concise knowledge models simplifies complex problems and facilitates handling large rule bases effectively. The proposed architecture supports the separation of the inference mechanisms from the knowledge base and utilises the meta-level representation to capture the problem-solving and decision making knowledge. The proposed techniques perform very well in diagnostic reasoning. In particular, the hierarchical one is capable of diagnosing multiple faults