基于变量转换法的混合离散变量优化设计

Optimization Design of Mixed Discrete Variable based on Variable Transformation Method

分析混合离散变量采用的优化算法的不足,针对混合离散变量优化提出变量转化法,即设定统一连续的名义变量,使数学模型可以调用连续变量的算法或软件,编写变量转化程序,将名义变量转化为实际离散变量,代入目标函数和约束条件,得到最优解后不需再进行变量圆整,保证了优化的准确性。以直齿圆柱齿轮传动优化设计为例,得到最优解并与其他常用的优化方法比较。结果表明,变量转换法简便易行,效果优于其他常用算法。

The deficiency of mixed discrete variables for the current optimal design is analyzed.A variable transformation method for mixed discrete variables is proposed,that is nominal variables are set,and software or algorithm of continuous variables could be called for mathematical model.The variables transformation program is written according to their relationship with real variables.Then,the nominal variables are put into the objective function and constraint condition,and the optimal solution is got.The variables of the optimal solution and real variables have the same type,so,it needn＇t to be rounded and ensured the accuracy of the optimization.The variable transformation method is applied into optimization design of spur gear transmission,and compared with the existent ordinary optimization methods.The obtained results show that the variable transformation method is simple and feasible and the optimum solution of the variable transformation method is superior to solutions of other algorithm.