Misfire Detection of Diesel Engine Based on Artificial Neural Networks

Engine misfire can be detected through on-board diagnostics (OBD) by means of instantaneous crankshaft angular acceleration analyzing and threshold-based classification rules. This method has certain limitations in diagnosing single-cylinder complete misfire conditions when the internal combustion engine is running at high speed and light load. The amplitude-frequency and phase-frequency characteristics of the instantaneous crankshaft rotational speed under misfire and normal operating conditions are compared. By extracting the amplitude and phase information of different harmonics and combining the artificial neural network (ANN) as a fault pattern recognition tool, an approach is proposed. Through bench experiments, the improved method is tested for fault diagnosis of complete misfire of single cylinder, complete misfire of two cylinders and certain degree of misfire of single cylinder at different speed on a diesel engine test-rig. The results show that the method can effectively identify different misfire modes under experimental conditions, and identify the misfire severity and cylinder location in the single-cylinder misfire mode with high accuracy. At the same time, the method can realize the misfire diagnosis within a certain speed range by training the neural network through a small amount of working condition data. It has strong feasibility and is expected to be used for the on-line diagnosis of engine misfire faults. © 2020, Editorial Department of JVMD. All right reserved.