| Summary: | This thesis presents an automation framework for the formulation of knowledge models to support the generation of explanations for engineering systems that are represented by the resulting models. Such models are assembled from instantiated generic component descriptions, known as model fragments. The model fragments are of sufficient details that generally satisfy the requirements of information content as identified by the user asking for explanations. Using a combination of Bayesian Networks and Approximate Reasoning techniques, in order to cope with different types of uncertainty arising from these requirements, model fragments are selected from a library and they are assembled prior to extraction of any textual information upon which to base the explanations. The thesis proposes and examines the techniques that support the fragment selection mechanism and the assembly of these fragments into models. It also addresses the issues concerning the scalability of the approach taken, with respect to a large physical domains.
|
|---|
