槽腔耦合液体机械密封多目标优化研究

发布时间：2018-12-09 14:21

【摘要】：机械密封端面造型及性能优化是当前机械密封研究和应用领域的热点课题。具有螺旋槽和微凹腔耦合端面的非接触式机械密封,是兼顾泵送效应和动压效应而获得优秀密封性和润滑性的良好密封装置。对于槽腔耦合机械密封的优化研究,涉及形貌组合、多参数、多目标等问题,目前这方面的研究还较少。本文在国家自然科学基金项目(51279067)的资助下,以槽腔耦合液体机械密封为对象,基于多元回归分析、神经网络和智能算法对其进行多目标优化研究,以期为机械密封的研究、优化和设计提供参考依据。主要研究工作及结论如下:1.基于空化模型和运用动网格技术,对4种不同槽腔组合方式的机械密封内流场进行数值模拟计算。通过对不同方案内流场静压分布、液膜刚度、摩擦扭矩和泄漏量的对比分析,获得了槽腔耦合机械密封端面造型的最佳组合方式,即当螺旋槽开在动环内径侧、微凹腔开在动环密封坝区时,能够获得更优的密封性能,为槽腔耦合优化提供基础。2.通过分析比较选取槽腔耦合机械密封的槽深h、螺旋角α、腔深h_p、面积比S_p为优化变量,选取液膜刚度K、泄漏量Q为优化目标,建立均匀试验表格U_(50)(50~4),利用多元回归分析法获得变量与目标之间的拟合方程,建立两者之间的近似模型,并利用MATLAB作出响应等高图,得到结构参数交互影响时对密封性能的影响规律,以及最优参数组合的取值范围。3.联合神经网络和智能算法建立了适用于槽腔耦合液体机械密封的多目标优化策略。首先利用MATLAB神经网络工具箱构建BP神经网络并训练样本对,获得变量与目标之间的适应度函数,再利用优化工具箱对适应度函数进行多目标寻优。通过遗传算法内循环和预测精度检查校正循环,将次优结果作为新样本加入到训练中,以保证预测精度,并通过不断调整得到密封几何参数的Pareto前沿解集。然后分别从多元回归近似模型和神经网络预测模型得到的Pareto中选取两组最优解方案进行优化前后的对比,结果表明优化方案opt1和opt2的液膜刚度分别比原始方案增大了12.58%、15.47%,泄漏量分别降低了26.56%、27.91%。4.采用激光加工技术对优化前后的密封环进行表面造型加工,借助MSTS-IV机械密封试验台进行密封性能试验,验证了优化结果的正确性。
[Abstract]:The modeling and performance optimization of mechanical seal face is a hot topic in the field of mechanical seal research and application. The non-contact mechanical seal with the coupling face of spiral groove and microcavity is a good sealing device which can obtain excellent sealing and lubricity by taking both pump effect and dynamic pressure effect into account. The research on the optimization of slot cavity coupling mechanical seal involves many problems, such as morphology combination, multi-parameter, multi-objective, etc. At present, there are few researches on this aspect. This paper, supported by the National Natural Science Foundation of China (51279067), takes the slot cavity coupling liquid mechanical seal as an object, and based on multivariate regression analysis, neural network and intelligent algorithm to study its multi-objective optimization, in order to study the mechanical seal. Optimization and design provide reference basis. The main research work and conclusions are as follows: 1. Based on cavitation model and dynamic grid technology, the numerical simulation of the internal flow field of four different groove cavity combinations is carried out. Through the comparison and analysis of hydrostatic pressure distribution, film stiffness, friction torque and leakage rate in different schemes, the best combination method for forming the end surface of mechanical seal with slot cavity coupling is obtained, that is, when the spiral groove is open in the inner diameter of the moving ring, When the microconcave cavity is opened in the dynamic ring seal dam area, it can obtain better sealing performance, which provides the foundation for the groove cavity coupling optimization. 2. By analyzing and comparing the slot depth h, spiral angle 伪, cavity depth hp, area ratio SSP as the optimization variable, the liquid membrane stiffness K and the leakage quantity Q are selected as the optimization targets. The uniform test table U50 (504) was established. The fitting equation between the variables and the target was obtained by multiple regression analysis. The approximate model between the two was established, and the response contour diagram was made by using MATLAB. The effects of structural parameters on sealing performance and the range of optimal parameters combination are obtained. Combined with neural network and intelligent algorithm, a multi-objective optimization strategy for slot cavity coupling liquid mechanical seal is established. Firstly, the MATLAB neural network toolbox is used to construct the BP neural network and to train the sample pairs to obtain the fitness function between the variables and the target. Then, the fitness function is optimized by using the optimization toolbox. In order to ensure the prediction accuracy, the sub-optimal results are added to the training by the genetic algorithm inner cycle and the precision check correction cycle, and the Pareto frontier solution set of sealed geometric parameters is obtained by adjusting continuously. Then, two groups of optimal solution schemes were selected from the multivariate regression approximation model and neural network prediction model to compare before and after optimization. The results showed that the liquid membrane stiffness of opt1 and opt2 were 12.58% higher than that of the original scheme, respectively. 15.47, the amount of leakage was reduced by 26.56% and 27.91%, respectively. The seal ring before and after optimization was machined by laser processing technology, and the seal performance was tested by MSTS-IV mechanical seal test rig, which verified the correctness of the optimized results.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH136

【参考文献】