基于代理模型的安全阀动态性能优化  

Optimization of Dynamic Performance of Spring-loaded Safety Valves Based on Agent Modeling

作　　者：谢鼎盛 周泽余 赵鑫宇 张斐然 石茂林 宗超勇 XIE Ding-sheng;ZHOU Ze-yu;ZHAO Xin-yu;ZHANG Fei-ran;SHI Mao-lin;ZONG Chao-yong(College of Electical and Mechanical Engineering,Harbin Engineering University,Harbin I50O0l,Heilongiiang,China;Wuhan Second Ship Design and Research Institute,Wuhan 430205,Hubei,China;School of Mechanical Engineering,Jiangsu University,Zhenjiang 212013,Jiangsu,China)

机构地区：[1]哈尔滨工程大学机电工程学院,黑龙江哈尔滨150001 [2]武汉第二船舶设计研究院,湖北武汉430205 [3]江苏大学机械工程学院,江苏镇江212013

出　　处：《阀门》2024年第9期1097-1102,共6页Chinese Journal of Valve

基　　金：国家自然科学基金项目(52205251);中央高校基础科研业务费(3072024CFJ0703)。

摘　　要：针对弹簧式安全阀在动态过程中可能存在启闭压差大、动态非稳定等问题,本文提出一种基于多输出最小二乘法支持向量机(Multi-output Least-Squares Support Vector Regressor, MLS-SVR)的安全阀动态性能代理模型,通过灰狼多目标算法实现阀门动态性能的综合优化。结果表明:MLS-SVR代理模型精度优于单目标支持向量机(Support Vector Regressor, SVR)模型;经灰狼多目标算法优化后,在稳定系数达到1.03时,安全阀的启闭压差由14.54%降为3.98%。该结果验证了MLS-SVR在安全阀动态性能优化方面具有良好的适用性,且灰狼多目标优化算法相较于单目标优化具有更好的综合性能。Aiming at the problems of large opening and closing pressure difference and dynamic instability in the dynamic process of spring safety valve, this paper proposes a dynamic performance proxy model of safety valve based on Multi-output Least-Squares Support Vector Regressor(MLS-SVR).The comprehensive optimization of valve dynamic performance is realized by grey wolf multi-objective algorithm.The results show that the accuracy of the MLS-SVR agent model is better than that of the single-objective Support Vector Regressor(SVR) model;after optimization by the Gray Wolf multi-objective algorithm, the opening and closing differential pressure of the safety valve is reduced from 14.54% to 3.98% when the stability coefficient reaches 1.03.The results verify that MLS-SVR has good applicability in optimizing the dynamic performance of safety valves, and the Gray Wolf multi-objective optimization algorithm has better comprehensive performance than single-objective optimization.

关 键 词：弹簧式安全阀 代理模型 多目标优化 多输出最小二乘法支持向量机 

分 类 号：TH134[机械工程—机械制造及自动化]