A Fault Diagnosis Method for Valve Train of Diesel Engine Considering Incomplete Feature Set

Abstract

Abnormal valve clearance is a common fault of diesel engine, and early warning of abnormal valve clearance plays an important role in the condition based maintenance of diesel engine. Although information fusion technology can improve the accuracy of fault diagnosis, it cannot guarantee that the fused features can perfectly represent the required key information. For the incomplete feature set, a method combining multi-domain feature and improved support vector machine is proposed. Firstly, the extraction of multi-domain feature is carried out to deeply explore the state information of valve train contained in the original vibration signal. The statistical characteristics and waveform characteristics are extracted from time domain vibration signals, and the frequency domain feature similar to time-domain feature is extracted after the Fourier transform of the vibration signal. What’s more, according to the working principle of diesel engine, the energy characteristics in angular frequency domain are extracted. Then, an improved support vector machine method based on multi-domain feature is proposed to further reduce the diagnostic errors caused by incomplete feature set. Finally, the proposed method is compared with other traditional methods about the fault diagnosis of valve train of diesel engine. The results show that the proposed method is applicable to the fault diagnosis of valve train of diesel engine with good accuracy, and the generalization ability of diagnostic model is greatly improved.