Knowledge is often structured hierarchically and decision rules are considered as mapping between hierarchically structured evidences and hypotheses. Evidence is gathered at various levels of the evidence hierarchy and conclusions are required at various levels of the hypothesis hierarchy. In this thesis, an inference method is proposed to draw a conclusion at any required hypothesis level based on the evidence gathered at various levels.
In the proposed inference scheme, an evidence is propagated up or down in the hierarchy to search the most suitable inference path to the requested hypothesis hierarchy. Evidence propagation, both within a hierarchy and between the hierarchies, are formulated in terms of fuzzy logic. Developed inference scheme has been applied to the development of an expert system which recommends welding process based on the various parameters such as materials, shape of juncture, location, etc.