Neuro-fuzzy-based fault detection of the airflow sensor of an idling gasoline engine

This paper presents a neuro-fuzzy-based diagnostic system for detecting the faults of an air flow sensor of an idling gasoline engine. Based on the Takagi-Sugeno fuzzy system model, the diagnostic system is formulated by the input-output relationships between symptoms and faults. The system parameters are regulated with the learning datum from experiments, using the steepest descent method and back-propagation algorithm. The proposed diagnostic system consists of two parts: one is to judge the fault of the sensor and the other is to identify the bias degree of the sensor. The experimental results show that the fault source and fault (bias) degree can be identified with the proposed diagnostic system, and indicate that the neuro-fuzzy strategy is an efficient and available approach for fault diagnosis problems of the gasoline engine system.