动态惩罚函数非线性规划遗传算法及在汽车变速器中的应用  

Dynamic Penalty Function Nonlinear Programming Genetic Algorithm and the Application in Automobile Gearbox

作　　者：闫帅印 薄瑞峰[1,2] 李瑞琴[1] 阴书玉 沈兴全[1,2] 潘红军[1,2] 

机构地区：[1]中北大学机械与动力工程学院,山西太原030051 [2]山西省深孔加工工程技术研究中心,山西太原030051

出　　处：《机械传动》2015年第2期146-149,共4页Journal of Mechanical Transmission

基　　金：国家自然科学基金(51275486);高等学校博士学科点专项科研基金(20111420110005);山西省回国留学人员科研项目(2014-050)

摘　　要：针对传统遗传算法在求解非线性规划问题时局部搜索能力较弱,惩罚函数求解精度不高的缺陷,将非线性规划算法引入到遗传算法中,提出一种基于动态惩罚函数的非线性规划遗传算法,将遗传算法的全局寻优能力和非线性规划算法的局部寻优能力结合起来,并引入动态惩罚函数,根据不可行点到可行域的距离和可行度自适应的调整惩罚项的值,从而能够快速求出全局最优解。介绍了动态惩罚函数的设计、改进遗传算法的关键技术和流程。最后,以某型号汽车变速器的优化设计验证了算法的合理性。与传统遗传算法相比,改进后的遗传算法解的质量、收敛速度明显提高,因而为遗传算法的改进提供了一种新的思路。Aiming at the defects of weak local search ability and the low solution accuracy of penal- ty {unction when solve the nonlinear programming problem, the nonlinear programming algorithm is in- troduced to the genetic algorithm and a nonlinear programming genetic algorithm is proposed based on dynamic penalty function. Combining the capable of global optimization of the genetic algorithm and the capable of local optimization and introducing dynamic penalty function, according to the value of penalty term is modified adaptively based on the distance of infeasible points to feasible solution space and feasibility degree, the global optimal solution is quickly to calculate. The design of dynamic penalty function and the key technology and process of improved genetic algorithms are introduced. Finally, the reasonability of algorithm is verified based on the example of the optimum design of a certain automo- bile gearbox. Compared with the traditional genetic algorithm, the solution quality and converged speed of improved genetic algorithms are improved obviously. As a result, a new way of thoughts is provided for genetic algorithm improvement.

关 键 词：动态惩罚函数 非线性规划遗传算法 局部最优解 全局最优解 

分 类 号：U463.212[机械工程—车辆工程] TP18[交通运输工程—载运工具运用工程]