A method for fusing multiple sources for noise reduction in a strong noise environment

To address the problem that existing fault diagnosis methods are unable to solve the problem of fault identification being significantly affected by the integration of multi-source sensor information, which is interspersed with varying degrees of noise, a model applicable to noise reduction of multi-source information is proposed, which utilises the global feature extraction capability of Transformer and the local feature extraction capability of the Parameter-Free attention module(SimAM) to achieve noise reduction. The experimental results show that the designed model can significantly improve the accuracy of multidimensional data fault classification in a strong noise environment.