A Fault Diagnosis Method for Fuel Injectors Using Machine Sound

Construction machine diagnostics based on engine sound is very important to enable a class of self-diagnosis applications that are very essential in the future autonomous systems. In this paper, we propose a fault diagnosis method to detect the abnormality of fuel injectors on the basis of machine sound. Analyzed results show that some frequency components appear in an abnormal engine sound that do not exist in a normal one. By comparing the average level of those frequency components to that of the explosion frequencies of the engine sound, an abnormal engine sound can be detected. Experiments are conducted to validate the analytical results. Moreover, in the event of abnormal sounds, principal component analysis (PCA) is employed to identify which injectors are faulty. The experimental results also show that fault injectors can be detected by PCA with high accuracy.