Study on Fault Diagnosis for Engine Based on Feature Fusion

Abstract

Fault Diagnosis for engine is very crucial, yet very difficult. Because vibrator-based techniques have been proven to be effective in detecting faults in rotating machine, the paper presents a new method for engine fault diagnosis by using the rotor's vibration signals. The experiment results show that the proposed method has prominent classification performance. First, to include more fault information, two kinds of features, i.e., spectrum features and wavelet features, are extracted from vibration signals of engine's rotor, respectively. Subsequently, two kinds of features are combined and PCA is further used to reduce the dimension of the combined features and remove their redundancy, thus we can obtain the fused features. Finally, the fused features are sent to BP network to accomplish the diagnosis of the faults in engine. The performance of the proposed method is tested, and it shows that the method can significantly improve classification ability.