Failures in basements are one of the most common problems in residential buildings in the United States. An expert frequently is needed to assess the causes of basement failures. Hiring an expert to diagnose the cause of the damage of one single house is often more expensive than repair the cracking caused by the basement failure. For this reason, many owners shy away from hiring consultants to inspect the condition of the basement.;Although structural and soil mechanics theory is used to discover the causes of basement failures, experts frequently use intuition, rules of thumb, and heuristic approaches. Despite the fact that these approaches often are not quantifiable, an excellent and affordable engineering tool could be constructed if the opinions of these experts were to be incorporated into a computerized system.;The result of this dissertation is the development of the Basement Failure Diagnosis Expert System, BAFDES, an expert system that can identify causes of basement failure. BAFDES is very user friendly, provides on-screen help and explanations at the user's request, and indicates how the conclusions of a consultation are reached.;To implement BAFDES, this dissertation introduces an innovative and effective method for the production of semantic net models based on fault trees. Semantic net models are then translated into a classificatory tree. The method accomplishes this through three steps that utilize the caused-by arcs in the semantic net, identify tests, and create intermediate branching by grouping alternatives in the classificatory tree. BAFDES has been tested and validated by the knowledge engineer and his adviser, by the participating expert, and by non-participating experts. The result indicates that the system completeness, user interface, and efficiency of BAFDES range between good and very good. The applicability of BAFDES ranges between fair and good.
展开▼
