Fault diagnosis of spent fuel shearing machines based on improved residual network

A spent fuel shearing machine is an important piece of equipment in the spent fuel reprocessing plant of power reactors, and it is important to carry out fault diagnosis of spent fuel shearing machines. The residual network has achieved some success in the fault diagnosis field, but there are currently two shortcomings. First, When the residual network is confronted with a large number of strong background noise signals, its learning ability decreases, leading to a decrease in diagnostic accuracy. Second, the finiteness and high information content of the sample data lead to an unsatisfactory diagnostic effect. In this study, an improved residual network fault diagnosis model for spent fuel shearing machines is constructed. The residual network is improved based on soft thresholding and pyramid squeezing attention to achieve the improvement goal of filtering background noise and learning richer multi-scale feature information to fully utilize the information in the data. Experimental validation and comparative analysis are performed using the operating noise data of the spent fuel shearing machines. The experimental results show that the proposed model can accurately recognize the fault categories of shearing machines, and there is a significant improvement in the accuracy rate compared with other methods.

Automated summary

An improved residual network fault diagnosis model for spent fuel shearing machines is proposed in this study, which achieves higher diagnostic accuracy by improving the residual network with soft thresholding and pyramid squeezing attention, and fully utilizes multi-scale feature information.