A damage diagnosis expert system with an analogical reasoning scheme embedded in a conventional deductive reasoning mechanism is proposed. In this system, the inference mechanism is implemented basically by a deductive inference mechanism (DIM) with forward chaining, while an analogical inference mechanism (AIM) is invoked when the DIM cannot achieve a diagnosis goal directly. The degree of similarity between two fuzzy truth values is evaluated according to an extended concept of similarity relation and used to guide the analogical inference. It is novel in at least two aspects: (1) A case base is introduced for knowledge representation. (2) The analogical inference technique is applied for the inference when sufficient experience is lacking to obtain reliable rules.