Intelligent Fault Diagnosis of Rotating Machinery Based on Acoustic Field Signal and Domain Adaptation

Equipment to perform field tasks exist in high humidity, high temperature, high pressure, high corrosion and other equipment non-contact situation, the relevant accompanying security is not suitable for the installation of contact sensors to collect the state signal of the equipment, and field complex task profile equipment is often in a complex variable condition state. Therefore, it is necessary to use non-contact sensing means to obtain the state information of the equipment, and adopt monitoring, diagnosis and prediction methods applicable to complex variable working conditions to conduct online real-time evaluation and accurate diagnosis and prediction of the technical state and health status of the equipment. Based on this, an intelligent diagnosis method for rotating machinery based on acoustic field signal and domain adaptation is proposed. The method first uses wavelet transform to noise reduction of the sound field signal containing strong background noise, and obtains wavelet map. Then, the wavelet map features are extracted using residual neural network to achieve source-domain fault diagnosis. Also, the joint maximum mean difference is used to achieve migration diagnosis under different working conditions. This method can effectively solve the problem that traditional contact sensors cannot adapt to the complex environment and the variable operating conditions diagnosis in actual operation.