Methodology of diagnostic knowledge acquisition from databases of examples

The diagnostics of a critical machinery is nowadays aided by expert systems. Their knowledge bases contain knowledge on diagnostic relationships between a technical state of a given machine, its operating conditions and observable symptoms of this state. This knowledge may be acquired either from human experts or from databases containing examples. The paper focuses on problems of knowledge acquisition from examples. It concerns the whole range of problems starting from preparation of examples up to the verification and validation of the knowledge base at the end of the proceeding. To acquire diagnostic knowledge from a set of examples we apply both machine learning and knowledge discovery methods. An example of the application of described methods for the acquisition of knowledge suitable for diagnosing complex technical states of rotating machinery is also given.