Wind Turbine Drivetrain Gearbox Fault Diagnosis Using Information Fusion on Vibration and Current Signals

Abstract

To improve the reliability of the conventional vibration-based wind turbine drivetrain gearbox fault diagnosis system, this article proposes a novel fault diagnosis method by fusing the information from gearbox vibration and generator current signals. First, the fault features contained in the gearbox vibration signals and the generator current signals are analyzed, respectively. Second, a multiclass support vector machine (SVM) model with probabilistic output is proposed to design two classifiers which output the probabilities of different gearbox fault types according to the input fault features extracted from the vibration signals and the current signals separately. Then, a nontrainable combiner and a trainable combiner are designed based on the Dempster-Shafer theory and the softmax regression technique, respectively, to fuse the information from the vibration and current SVM classifiers at decision level. The output of each combiner is the final diagnosis result. The proposed method is validated by experimental results obtained from a test gearbox with different types of faults. The validation results show that the proposed method can increase the fault diagnostic accuracy and is more robust than the conventional fault diagnosis systems that only use one type of signals.