Modeling method embedded into diagnostics, reliability and maintenance - models as knowledge representation systems

The paper introduces the problem of building diagnostic models according to knowledge engineering conceptualization and methodology, where the model is a formal and computable symbolic representation of the specific domain knowledge. Diagnostic process is conceptualized as computable inferential reasoning from observable symptoms to unobservable causes. Uncertainty inherent in such reasoning is represented with Bayesian networks technology. That methodological background of the diagnostic process modeling makes the model suitable for embedding into the physical system, adaptively fitting to specific operating conditions using learning algorithms on the collected data, and operating as the knowledge base answering the diagnostic questions under given observations. Presented problem conceptualization is exemplified with car ignition system. Presented example is minimal and sufficient to show the method of building Bayesian network learning conditional distributions and operating model by answering diagnostic questions.