基于BO-RF回归预测的海水柱塞泵配流阀结构参数优化研究  

Distribution valve of structural parameter optimization for seawater pistonpump based on BO-RF regression prediction

作　　者：周广金 国凯 孙杰[1,2] 黄晓明 ZHOU Guangjin;GUO Kai;SUN Jie;HUANG Xiaoming(School of Mechanical Engineering,Shandong University,Jinan 250061,China;Key Laboratory of High Efficiency and Clean Mechanical Manufacture of Ministry of Education,Shandong University,Jinan 250061,China;Mechatronics Engineering Department,Shandong Aeronautical University,Binzhou 256603,China)

机构地区：[1]山东大学机械工程学院,山东济南250061 [2]山东大学高效洁净机械制造教育部重点实验室,山东济南250061 [3]山东航空学院机电工程学院,山东滨州256603

出　　处：《机电工程》2025年第4期618-627,共10页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金青年基金资助项目(51605256)。

摘　　要：海水柱塞泵采用阀配流方式可以提高其密封性能,保证其具有较高的输出压力。针对配流阀结构参数设计不合理,导致阀芯运动滞后和容积效率降低的问题,提出了一种贝叶斯优化(BO)与随机森林算法(RF)相结合的海水柱塞泵配流阀结构参数优化方法。首先,利用AMESim软件搭建了海水泵液压系统仿真模型,利用试验验证了仿真模型的准确性,分别分析了吸、排液阀的弹簧刚度、弹簧预紧力、阀芯质量对阀芯滞后以及容积效率的影响;然后,基于仿真获得的配流阀结构参数与对应输出流量的数据,对比分析了贝叶斯优化随机森林(BO-RF)模型、粒子群优化随机森林(PSO-RF)模型、反向传播神经网络(BPNN)模型和随机森林(RF)模型的回归预测结果,以BO-RF模型为回归预测模型,利用遗传算法优化了配流阀结构参数,并获得了结构参数最优解;最后,对优化后的配流阀结构参数进行了仿真分析。研究结果表明:吸、排液阀的弹簧刚度、弹簧预紧力增大能够减小阀芯滞后,提高容积效率,参数增大到临界值后,容积效率会随参数增大而降低;吸、排液阀的阀芯质量增大会增大阀芯滞后,减小容积效率;BO-RF模型的均方根误差(RMSE)、平均绝对百分比误差(MAPE)、决定系数(R~2)均优于RF、PSO-RF和BPNN模型,其回归预测准确度更高;对于优化后的结果进行仿真可得:容积效率较原结构提高了4.7%。该模型适用于配流阀结构参数预测和优化问题,可为提高柱塞泵容积效率提供参考。Seawater piston pumps with valve distribution can improve sealing performance and ensure higher output pressure.Aiming at the problem of the motion lag of the valve core and the reduction of the volumetric efficiency,it caused by the unreasonable design of the structural parameters of the distributing valve,a method based on the combination of Bayesian optimization(BO)and random forest(RF)algorithm was proposed to optimize the structural parameters of the distributing valve of the seawater piston pump.Firstly,AMESim software was used to build a simulation model of the hydraulic system of the seawater pump.The accuracy of the simulation model was verified through experiments,and the effects of spring stiffness,spring preload,spool mass on spool hysteresis and volumetric efficiency of the suction valve and discharge valve were respectively analyzed.Then,based on the structural parameters of the distributing valve obtained from the simulation and the corresponding output flow rate,a comparative analysis of regression prediction results was performed on the Bayesian optimization random forest(BO-RF)model,particle swarm optimization random forest(PSO-RF)model,back propagation neural network(BPNN)model and random forest(RF)model.The BO-RF model was used as a regression prediction model to optimize the structural parameters of the distribution valve using genetic algorithm and the optimal solution of the structural parameters was obtained.Finally,simulation analysis was carried out for the optimized structural parameters of the distribution valve.The research results show that increasing the spring stiffness and spring preload of the suction valve and discharge valve can reduce the spool hysteresis and improve the volumetric efficiency.Still,volumetric efficiency is reduced after increasing the critical point.Increasing the spool mass of the suction valve and discharge valve can increase the spool hysteresis and reduce the volumetric efficiency.The root mean squared error(RMSE),mean absolute percentage error(MAPE),and c

关 键 词：三柱塞曲柄连杆式高压海水柱塞泵 容积效率降低 阀芯运动滞后 贝叶斯优化随机森林回归预测模型 粒子群优化随机森林 弹簧刚度和预紧力 阀芯质量 

分 类 号：TH137.51[机械工程—机械制造及自动化]