基于DOE和MIGA的消声器优化设计
Optimization Design of Muffler Based on Design of Experiment and Multi-Island Genetic Algorithm

高效率的设计出大消声量的消声器一直是车辆排气噪声控制中面临的难题。考虑到消声器优化过程中涉及参数较多,在消声器传递损失数值建模的基础上,采用试验设计(DOE)中的拉丁超立方设计对消声器参数进行分析,结合多岛遗传算法(MIGA)和传统遗传算法(GA)分别建立消声器在排气噪声单峰值频率和多峰值频率处的传递损失为目标的优化模型,开展消声器传递损失优化设计研究。结果表明:DOE方法能有效的辨识出各参数对消声器传递损失影响的大小,简化了消声器的优化模型。MIGA对消声器在单峰值频率和多峰值频率的优化都优于GA,且多峰值频率的优化好于单峰值频率的优化,能使排气噪声最大降低20.98 dB。
To design muffler efficiently with high capability of noise reduction is always a tough problem in control of vehicle exhaust noise. Considering that there are many parameters in the muffler design optimization, the design parameters of mufflers were analyzed by Latin Hypercube of design of experiment (DOE) based on the numerical modeling of transmission loss. Combining the multi-island genetic algorithm (MIGA) with genetic algorithm (GA), the optimization model of mufflers was established in which the transmission losses at the single peak frequency of exhaust noise and multiple peak frequencies are set as as the optimization objective, respectively. The result shows that the DOE method can effectively identify the parameters which affect muffler performance and simplify the optimization model of muffler. The optimization results of MIGA in both the single peak frequency and multiple peak frequency is better than that of GA, and the results of multiple peak frequency, optimization which is better than that of single peak frequency optimization, can reduce the exhaust noise by 20.98 dB. This study provides a new optimization design method of muffler