SVM和改进的PSO在压路机驾驶室结构噪声优化中的应用

Application of SVM and Improved PSO in Structure Noise Optimization for the Road Roller Cab

针对压路机驾驶室结构噪声,将拉丁超立方试验设计、支持向量机近似模型、改进的粒子群优化算法相结合,通过修改驾驶室主要板件的板厚参数降低压路机结构噪声。建立一套基于支持向量机和粒子群算法控制车内结构噪声的设计流程。针对粒子群可能出现局部最优解的问题,对粒子群进行了改进。并利用改进的粒子群优化支持向量机参数,构建高拟合精度的支持向量机模型代替有限元模型。并用改进的粒子群算法对该模型进行板厚寻优,找到一组最佳的板厚参数使得参考点(驾驶员右耳处)声压级最小,减少计算工作量,提高优化效率。

The Latin hypercube experimental design, support vector machine（SVM） approximation model and improved particle swarm optimization（PSO） algorithm were combined to reduce the structure noise of a road roller＇s cab by modifying the thickness parameters of the main plates of the cab. The design process for controlling the interior structure noise based on the SVM and PSO was presented. The particle swarm was improved by the PSO which could lead to a local optimal solution. The SVM model with high fitting accuracy was built and the improved particle swarm was used to optimize the plate thickness parameters of the SVM model. A set of the best parameters of the thickness that could minimize the sound pressure level at the reference point（the driver＇s right ear） was found. The process could save the computation workload and improve the efficiency of the optimization.