基于MCS和改进遗传算法的进气消声器优化分析  

Optimization Analysis of Acoustic and Resistance Characteristics of Intake Muffler Based on Monte Carlo Simulation

作　　者：朱传峰 毕嵘 韦静思 袁懋荣 李波 朱亚亚 ZHU Chuanfeng;BI Rong;WEI Jingsi;YUAN Maorong;LI Bo;ZHU Yaya(GAC Automotive Engineering Institute,Guangzhou 511434,China)

机构地区：[1]广汽集团广汽研究院,广东广州511434

出　　处：《现代车用动力》2020年第2期6-11,46,共7页Modern Vehicle Power

基　　金：国家重点研究计划资助项目(2017YFB0103300)

摘　　要：综合考虑发动机进气消声器声学性能和阻力特性,采用蒙特卡洛模拟(MCS),分别对基于传递矩阵和神经网络建立的进气消声器传递损失和压力损失数值模型进行参数贡献度分析,结合改进遗传算法(GA)对进气消声器进行单目标和多目标优化。研究结果表明:MCS方法有效辨识出参数L2,L4,L6,D2,D3,D4对传递损失和压力损失贡献都较大,简化了优化分析模型。基于神经网络建立的消声器压力损失数值模型精度较高,消声器压力损失大小的限制对进气消声器的优化结果影响较大。在满足压力损失的情况下,单目标优化能使进气消声器的传递损失在单个共振带中心频率处传递损失达到最大值,而多目标优化得到的进气消声器比原始进气消声器控制进气噪声最多降低5.31 dB,在整个工况范围,进气噪声基本都有所降低,性能优于单目标优化的结果。Considering the acoustic and resistance characteristics of intake muffler,the transfer matrix and neural network were used to construct the numerical calculation model of intake muffler transmission loss and pressure loss.The contribution of intake muffler parameters were analyzed based on Monte Carlo Simulation(MCS),combined with improved genetic algorithm(GA),the single objective and multiple objective optimization model were established respectively.The result shows that MCS can effective identification of parameters L2,L4,L6,D2,D3,D4 have great contribution to transmission loss and pressure loss,and simplify the optimization model.The precision of the intake muffler pressure loss model based on neural network is accurate and the limitation of pressure loss of intake muffler has great influence on the optimization.Under the condition of considering pressure loss,the transmission loss of intake muffler corresponding to the center frequency of single resonant band through single objective optimization can reach maximum,however,the multiobjective optimization is better than that of the original intake muffler to control the intake noise maximum reduction is 5.31 dB,and the performance of the intake muffler is better than that of single target optimization.

关 键 词：蒙特卡洛模拟 神经网络 遗传算法 传递损失 压力损失 

分 类 号：U464.134.4[机械工程—车辆工程]