基于神经网络的唇封结构优化方法

旋转唇形密封广泛用于机械行业，其密封圈的设计合理性直接影响着密封性能。采用有限元仿真、数值仿真等技术，研究唇口各个关键参数对密封圈性能的影响规律，并结合深度学习技术，利用Pytorch框架搭建出唇封性能神经网络预测模型，拟合出唇封各相关结构参数和其密封性能(泄漏率、摩擦力)之间的影响规律；利用建立的模型在参数空间中找到密封性能最优的一组参数值，通过经典唇封数值仿真技术验证该神经网络模型的准确性，提高了唇封结构优化设计效率。结果表明，基于Pytorch框架所搭建的非线性人工神经网络可建立精度较高的唇封结构参数和密封性能的非线性映射关系，利用这种方法可以快速找到优化程度更高的唇封结构参数，提高了唇封结构优化设计效率。

Optimization Method of Lip Seal Structure Based on Neural Network

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.