Prediction of spiral groove dry gas seal performance and intelligent optimization of structural parameters

Abstract

Using gas lubrication theory and finite difference methods, this study solve the gas film pressure equation in spiral groove dry gas seal and assesses key performance aspects like opening force, leakage rate, and gas film stiffness. It examines how structural parameters affect sealing performance and utilizes Latin hypercube sampling with a radial basis neural network, optimized by genetic algorithms, for predictive analysis. The effectiveness of control variable, orthogonal, and genetic algorithm methods in optimizing sealing performance is comparatively analyzed. Findings reveal a substantial impact of structural parameters on sealing performance with intricate interdependencies. The neural network, trained with 400 samples, achieves a prediction error below 3.6%. Genetic algorithm optimization of structural parameters notably enhances seal performance over other methods.