指尖密封结构参数的多目标混合教与学优化方法  

Multi-objective Hybrid Teaching-learning-based Optimization Method for Structural Parameters of Finger Seal

作　　者：王娟[1] 刘美红[1] 祝世兴[2] 陈文博 李遇贤[1] 孙军锋[1] WANG Juan;LIU Meihong;ZHU Shixing;CHEN Wenbo;LI Yuxian;SUN Junfeng(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500;School of Aeronautical Engineering,Civil Aviation University of China,Tianjin 300300;Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming,650500)

机构地区：[1]昆明理工大学机电工程学院,昆明650500 [2]中国民航大学航空工程学院,天津300300 [3]昆明理工大学信息工程与自动化学院,昆明650500

出　　处：《机械工程学报》2023年第9期157-170,共14页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(51765024)。

摘　　要：作为一种先进柔性密封,指尖密封结构参数对其性能的影响较为复杂,传统建模方法难以有效刻画两者间的映射关系,这导致指尖密封结构参数的优化难以实现。为此,首先利用BP神经网络构建迟滞率、平均接触压力和基圆半径、指尖梁个数等多个结构参数间的拟合关系,并在此基础上建立同时最小化迟滞率和平均接触压力(Minimize hysteresis-rate and average contact pressure,MHACP)的多目标优化问题模型;然后,设计结合Pareto支配法的混合教与学优化方法(Hybrid teaching-learning-based optimization,HTLBO)对所建模型进行求解;最后,通过仿真实验对BP神经网络、MHACP和HTLBO进行验证。结果表明:BP神经网络的整体线性回归拟合度超过0.99;95%置信区间(Confidence interval,CI)下,MHACP的仿真结果在等价区间[0.7,1]上与ANSYS等效,说明MHACP可以较好地反映指尖密封结构的特性,通过对MHACP的求解可以实现对指尖密封结构参数的优化;HTLBO具有较好的稳定性和优化性能,能同时有效改善指尖密封的迟滞和磨损问题并提供多组具有不同性能偏好的解以满足实际工程需求,为机械密封领域结构参数优化提供了一种普适方法。The finger seal is an advanced flexible seal,but the influence of its structural parameters on its performance is so complex that it is difficult for traditional modeling methods to optimize them for these methods are unable to describe the mapping relationship between structural parameters and its performance.To solve this problem,BP neural network is used to construct the fitting relationship among the hysteresis rate,the average contact pressure and the radius of the base circle,the number of finger beams and other structural parameters based on which a multi-objective optimization model is established to simultaneously minimize hysteresis-rate and average contact pressure(MHACP);Then the hybrid teaching-learning-based optimization(HTLBO)method is combined with Pareto domination method to solve the model;Finally,BP neural network,MHACP and HTLBO are verified by simulation experiments.The data indicated that the overall linear regression fit of the BP neural network exceeds 0.99;the simulation results of MHACP are equivalent to ANSYS on the equivalence interval[0.7,1]at 95%Confidence Interval(CI);and MHACP can better reflect the characteristics of finger seal structure and the optimization of finger seal structure parameters can be realized by solving MHACP.HTLBO,due to its good stability and optimization performance,can effectively improve the hysteresis and wearing of finger seals and provide multiple sets of solutions with different performance preferences to meet the actual engineering requirements.A universal method is provided for structural parameter optimization in the field of mechanical seals.

关 键 词：指尖密封结构 多目标优化 PARETO 教与学优化 

分 类 号：TG156[金属学及工艺—热处理]