Using prototypes in case based diagnosis of steam turbines

One of the requirements of diagnostic expert systems is the provision of an explanation of what they are doing and how they reach particular conclusions. This requires knowledge of the user in order to match his/her understanding, and also to be able to justify themselves like human experts do. When they are not understood, they should find alternative ways of explaining. This cannot readily be achieved by a single reasoning paradigm. The research reported in the paper deals with aspects such as multiple models and explanation. In fact, artificial intelligence (AI) paradigms such as rule, model or case based reasoning, have to work together. A general framework is proposed which includes different reasoning agents (e.g. model based reasoning (MBR), case based reasoning (CBR), rule-based systems (IKBS)), a control agent (Control), an explanation module (Explain), a Blackboard (BB) for exchanging and recording messages, and interfaces to domain libraries (Knowledge Bases (KB) e.g. models, cases, rules, etc.), databases (DB) (data from acquisition systems) and a user.