Optimization Research Of Upstream Pumping Mechanical Seal Based On Multi-objective Genetic Algorithms

Upstream pumping mechanical seal as a typical non-contact mechanical seal, because of its lubricating fluid, a small amount of leakage or no leakage, long life and other advantages, have attracted much attention, research and application.However, with the development of science and technology and sealing performance requirements continue to increase, how to more effectively design a good performance mechanical seal is an urgent need to study in depth. The shape parameters of mechanical seal face are the key factor to the merits of the sealing performance. Currently, simulation method is used in the optimization design of the parameters of the end face with a single factor or objective. Although there are certain effects, it is difficult to achieve high performance standards. Therefore, under funded by the National Natural Science Fund Project(project number: 51279067) and Aviation Science Fund Project(project number: 201328R3001), in this paper, the upstream pumping mechanical seal for the study, combined with neural networks and genetic algorithms to optimize multi-objective design studies. The main work and conclusions are as following:Based on cavitation model and dynamic mesh, there are to simulate micro gap inner flow field in upstream pumping mechanical seal and analyze the effect of end face geometry on the sealing performance. The study shows that the groove depth, spiral angle groove diameter ratio and groove width ratio are the main factors for the sealing performance; dynamic mesh can be effectively used in sealing the inner flow field.To obtain a function between the parameters of Surface Topography seal and sealing performance on upstream pumping mechanical, it uses a method of combining uniform design, multiple regression analysis and artificial neural network. The research shows that combining with uniform design and neural networks can be a good approximation of the true relationship between the geometric parameters and performance parameters;and there are interaction between geometric parameters and have an important impact on the sealing performance; the results of groove optimized interval: 6≤h≤10μm,16°≤α≤20°,0.35≤β≤0.55,0.4≤γ≤0.6.Based on multi-objective genetic algorithm, combining CFD numerical computation and neural network prediction model, it is established to the multi-objective genetic optimization strategy for the upstream pumping mechanical seal performance.The research shows:(1)Based on the reasonable design of experiment using the neural network method for forecasting model, it can obtain more accurate the relationship between optimization variables and the optimization goal,neural network method is more suitable for establish prediction model internal ambiguous relations;(2)The sealing performance optimized is improved, using multiple regression model liquid film stiffness increased by 7.5%, decrease the leakage by 16.6%, using the neural network prediction model and genetic algorithm to liquid film stiffness increased by 13.1%, decrease the leakage by 18.9%.Using MSTS-IV test bench for experimental verification optimization results of mechanical seals topography,the results showed that the test results are consistent with the simulation results,and the genetic algorithm optimization results is better.