改进的实数编码遗传算法在产品布局设计中的应用

Modified real-coded genetic algorithm and its application to layout design

为求解复杂结构产品的布局设计问题,对实数编码遗传算法进行了改进,直接将问题的求解变量作为染色体基因进行编码,提出了一种解空间编码遗传算法。在求解问题前,对布局问题进行了预处理。在求解过程中,首先引入模拟退火算法的思想对解进行选择;然后对不同类型变量采用不同的交叉和变异算子进行了处理,并在变异前对个体的连续变量随机加一个扰动量,以控制其搜索范围;最后对求解的中间结果按最优保留策略进行了处理。该算法实现了3维空间布局规划的自动寻优,并求解了2维和3维带性能约束的布局问题,验证了该算法的可行性和有效性。

To solve layout problems in complex structures of products, through improvements on the real-coded genetic algorithm, a Solution-Vector Coded Genetic Algorithm （SVCGA） was proposed. In the SVCGA, the variables of target function were regarded directly as genes of chromosomes and were coded. To solve layout design problems by SVCGA, the problem＇s pretreatment was necessary in the process of SVCGA evolution. Firstly, the idea of Simulated Annealing Algorithm （SAA） was introduced to select operator to control individual diversity. Then, different cross and mutating operators were acted on different variables. Moreover, to control the search range of the algorithm, a fluctuant scale was multiplied to the continuous variables before they were mutated. Finally, the result of each generation＇s evolution was managed according to the strategy of preserving the elitist. Through SVCGA,the automated spatial planning has been achieved in three-dimensional layout problems. In addition, to test the feasibility and effectiveness of SVCGA, a two-dimensional and a three-dimensional layout problem were introduced and solved by SVCGA. The layout results indicated that SVCGA was feasible and effective.