The invention provides a method for generating a model for judging the cause of an engine fault. The method comprises the following steps: A1) collecting various fault symptoms of an engine and at least one fault cause causing each fault symptom; A2) for each fault symptom, performing fuzzy processing on the fault symptom according to various fault degrees corresponding to the fault symptom to obtain a fuzzy fault symptom; A3) determining a fuzzy rule based on a corresponding relationship between the fuzzified fault symptom and the fault cause; and A4) training the fuzzy neural network according to the fuzzy rule to obtain a model for judging the cause of the engine fault, directly collecting some state parameters of the existing hardware and device of the engine, and analyzing whether thecorresponding fault symptom is about to occur or has occurred, thereby saving the labor cost.