Intelligent fault diagnosis of mechanical equipment based on industrial big data

Effective fault diagnosis will greatly improve the operational efficiency of industrial machinery and equipment. In this paper, for the issues of multi-fault coupling and low diagnostic accuracy that exist in the current research. Firstly, the mechanical equipment signals are pre-processed. The empirical modal decomposition is introduced to construct the fault eigenvectors of industrial mechanical equipment. Then the improved principal component analysis is used to map the high-dimensional features to the low-dimensional space, the dual attention mechanism (DAM) is introduced to improve the transformer model (ODAT), an ODAT model is trained for each fault for diagnosis, and a fault set is generated based on the diagnosis results of all ODAT models. Comparative experiments were conducted on the PHM15 dataset, and the results show that the fault diagnosis accuracy of the proposed model is 93.71%.