Compound Fault Diagnosis Framework based on Multi-Source Information Fusion

Condition monitoring and fault diagnosis of rotating machinery play an important role in the field of prognosis and health management (PHM). To address the challenges in compound fault diagnosis of rotating machinery, we propose a fusion method in feature-level for multi-sensor data, which can extract effective fault features from multi-sensor data. Meanwhile, we employ a multi-output classifier to solve the decoupling of compound faults which is able to recognize unseen compound fault combinations. Based on these two techniques, we designed a CNN model for compound fault diagnosis. In our model, the multi-sensor data are first fed into two parallel channels, time-domain and frequency-domain, where the features in multi-sensor data are extracted and fused in the time and frequency domains respectively, then the extracted features go into a module where the time-domain and frequency-domain features are fused, and finally, multi-output classifier is used to decouple the compound fault and output classification decision. The experimental results show that the proposed compound fault diagnosis model has achieved good performance.