Intelligent Diagnosis of Rolling Bearing Faults in Wind Turbines Based on Deep Network SelfLearning

With the rapid development of modern industry, the operating environment of large rolling bearing equipment is characterized by complex and changeable conditions, which makes the acquired data often contain noise, incomplete and unlabeled. Therefore, it is difficult for traditional pattern recognition-based methods to establish effective diagnosis model. An intelligent fault classification and diagnosis method based on deep belief network is proposed to solve this problem. This method uses incomplete data for modeling. Firstly, normalized the bearing fault signal, exploit fault-related feature representations by using the greedy learning of deep belief network (DBN) layer by layer, and the feature extracted by DBN is taken as the input of the fuzzy classifier. In the training stage, the traditional fuzzy C-means algorithm was used to train the classifier to identify the known faults and normal operating states. Finally, in the online application stage, an improved fuzzy C-means clustering algorithm is proposed, which is used to identify outliers in unknown data, and the non-outliers are classified by the classifier constructed in the training stage. The results show that the proposed method can not only accurately classify the fault type and damage degree, but also realize fault self-learning and fault library self-growth, and enhance the intelligence of classification process in case of noisy and incomplete data modeling.