A Fault Diagnosis of Rotating Machinery Based on a Mutual Dimensionless Index and a Convolution Neural Network

Abstract

For the fault diagnosis process of petrochemical rotating machinery, it is difficult to accurately identify faults by relying only on dimensionless index methods. Therefore, a fault diagnosis of rotating machinery based on mutual dimensionless index and convolution neural network is proposed. Firstly, collects the rotating machinery fault signal of the petrochemical large unit and mutual dimensionless index. Then, the sensitivity analysis of mutual dimensionless index, and carried out to extract the sensitive features. And then, the sensitive feature samples are mapped to the common subspace of the adversarial network for capacity augmentation. Finally, the sensitive features sample after capacity is input to the convolutional neural network for recognition. Through the verification of the petrochemical experimental platform fault and the wind turbine blade fault, The proposed method has a good diagnosis effect and can adapt to complex on-site conditions.