基于层级分解的前围声学包多目标优化  

Multi-objective optimization of front wall acoustic package based on hierarchical decomposition

作　　者：杨帅[1,2] 吴宪 薛顺达[2] YANG Shuai;WU Xian;XUE Shunda(School of Automotive Studies,Tongji University,Shanghai 201804,China;SINETAC Auto(Shanghai)Co.,Ltd.,Shanghai 201100,China)

机构地区：[1]同济大学汽车学院,上海201804 [2]拾音汽车科技(上海)有限公司,上海201100

出　　处：《振动与冲击》2025年第3期267-277,共11页Journal of Vibration and Shock

摘　　要：搭建了前围声学包多层级目标分解架构,提出GAPSO-RBFNN(genetic algorithm particle swarm optimization-radial basis function neural network)预测模型,并将其应用于多层级目标分解架构。将材料数据库、覆盖率、泄漏量作为优化的变量范围,以PBNR(power based noise reduction)均值作为约束,以质量和成本作为优化目标,采用非支配排序遗传算法(nondominated sorting genetic algorithm II,NSGA-II)进行多目标优化,得到Pareto多目标解集。并从中选取满足设计目标的最佳组合方案(材料组合、覆盖率、前围过孔密封方案选型)。结果显示,该模型最终的优化结果与实测结果接近,误差分别为0.35%,1.47%,1.82%,相较于初始声学包方案,优化后的结果显示,PBNR均值提升3.05%,其质量降低52.38%,成本降低15.15%,验证了所提方法的有效性和准确性。Here,a multi-level target decomposition architecture for front wall acoustic package was built,a genetic algorithm particle swarm optimization-radial basis function neural network(GAPSO-RBFNN)prediction model was proposed,and it was used in multi-level target decomposition architecture.Material database,coverage,and leakage were taken as optimization variable ranges,PBNR(power-based noise reduction)mean was taken as constraints,weight and cost were taken as optimization objectives,and NSGA-II(nondominated sorting genetic algorithm II)was used to perform multi-objective optimization,and obtain Pareto multi-objective solution set,from which the best combination scheme including material combination,coverage and front wall through-hole sealing scheme selection was chosen to satisfy design objectives.The results showed that the final optimized results of the proposed model are close to the measured results,errors are 0.35%,1.47%and 1.82%,respectively;compared with the initial acoustic package scheme,after optimization,the average PBNR for front wall acoustic package increases by 3.05%,its weight decreases by 52.38%,and its cost decreases by 15.15%;the effectiveness and correctness of the proposed method are verified.

关 键 词：GAPSO-RBFNN 声学包 PBNR NSGA-II Pareto多目标解集 

分 类 号：U27[机械工程—车辆工程] U46[交通运输工程—载运工具运用工程]