This paper describes a fuzzy model that is designed to diagnose automotive engineering faults. The fuzzy model has two modes, L-mode, which is the fuzzy learning mode and T-mode, which is the test mode. In the L-mode, the system learns two types of engineering diagnostic knowledge, expert knowledge, and the knowledge acquired from training data using machine learning techniques. A fuzzy diagnostic system for engine vacuum leak detection has been implemented based on this fuzzy model. The system has been tested on the data downloaded directly from the test sites of assembly plants of the Ford Motor Company, and its performance is excellent.