Rotary lip seals are widely used in the mechanical industry,and the rationality of their sealing ring design directly affects the sealing performance.The influence of key parameters of the lip on the performance of the sealing ring was studied using finite element simulation,numerical simulation and other techniques.Combined with deep learning technology,a neural network prediction model for lip sealing performance was constructed using the Pytorch framework.The influence law between the relevant structural parameters of the lip seal and its sealing performance (leakage rate,friction force) was fitted,and the model constructed was used to find the optimal set of parameter values for sealing performance in the parameter space.The accuracy of the neural network model was verified through classical lip sealing numerical simulation technology,which improved the efficiency of lip sealing structure optimization design.The results show that the nonlinear artificial neural network built based on the Pytorch framework can establish a high-precision nonlinear mapping relationship between lip seal structural parameters and sealing performance.This method can quickly find more optimized lip seal structural parameters,improving the efficiency of lip seal structural optimization design.