Intelligent Faults Diagnosis of Variable Operating Condition Bearing Based on Order Analysis and Deep Residual Networks
Chenchen Song,Jin Li,Xiang Zhou,Yafeng Wu
DOI: https://doi.org/10.1109/sdpc62810.2024.10707711
2024-01-01
Abstract:As an important component of rotating mechanical systems, rolling bearings greatly affect the safety and reliability of equipment operation. If rolling bearings fail during system operation, it will not only lead to a decline in equipment performance but also, in extreme cases, cause system failure or shutdown, resulting in more serious economic losses or even casualties. Therefore, fault diagnosis of rolling bearings is of great significance. However, due to the complex working environment, accurate fault diagnosis of rolling bearings faces significant challenges. For example, in reality, there are often variable operating conditions, and changes in equipment speed can lead to signal “spectrum damage,” making it difficult to obtain effective information for fault identification and extraction. This ultimately makes it difficult to identify bearing faults. This research proposes an order analysis and deep residual network based fault diagnosis technique for rolling bearings working under changeable conditions. Firstly, the original vibration signal and speed signal data of the bearing are preprocessed using order analysis method to generate the corresponding target data set. Then, the deep residual network model is trained and fine-tuned on the generated target data set. Finally, the fine-tuned deep residual network model is applied to fault diagnosis. The method is validated on a variable operating condition bearing data set from the University of Ottawa in Canada, and the results show that the method can achieve a 99.4% accuracy rate in fault diagnosis of variable operating condition bearings, demonstrating promising application prospects.