均匀设计与遗传算法相结合已有许多成功例子,但在应用中均匀设计表一般囿以固定形式而使二者的结合缺乏灵活性和普适性.为此提出应用亚遗传算法获得若干个任意水平数的均匀设计,并将其以3种方式嵌入标准遗传算法中:1)初始种群的均匀性分布;2)进化过程中对变量空间均匀性投点搜索;3)应用均匀设计进行调优试验,同时还增加了正态随机和摄动调优等试验操作技术,以上形成了基于试验设计、具有自适应能力的试验遗传算法.算例表明,通过以上方法改进的遗传算法可较好地保持种群多样性,寻优效率有较大提高,并能自动适应算法对搜索精度的要求.
Genetic algorithms (GAs) combined successfully with uniform designs (UDs) can be often found in many papers. Based on a stationary uniform table, however, most of these hybrid methods are lack of convenience and general applicability. So a new method using sub-GA to generate many a uniform designs with any levels was here introduced. The obtained UDs were imbed into GAs in three ways: 1 ) uniform generation of forerunner individuals; 2 ) uniform searching in variables space; 3 ) utilizing uniform designs to make evolution experiment. Meanwhile techniques of normal random distribution and variables perturbation were added into the searching process. So the socalled adaptive experimental genetic algorithms (AEGA) based on UDs was formed. Results of several examples showed that, the improved algorithms could keep the individual variety better, owning good characters of fine optimizing efficiency and precision adaptivity.