改进遗传算法在红外增透膜系设计中的应用

Application of Improved Genetic Algorithms for the Design of the Infrared Antireflection Coatings

在3～5μm红外增透膜系的遗传算法优化设计中，当群体接近成熟时，个体间的竞争能力和算法的区别选优能力会下降，进化将停止不前，产生无意义的重复计算和长时间等待。为了监控进化过程中染色体的变化趋势和改进情况，并提高算法的搜索能力，提出了一种采用离线比较法和辅助优选因子的薄膜改进遗传算法。计算实验表明，离线比较监控法能有效地掌握计算的进展情况，准确地监察到群体的成熟收敛、进化停止，并可据此决定改进算法的时机。采取辅助优选因子，能保持先前的计算成果，选出当前群体中的优良个体，跟引入的新鲜个体一同参与交叉，使进化不断向前，从而使改进后的算法搜到合格解的几率大大提高。

In the optimization design of infrared antireflection optical thin films for 3-5μm by genetic algorithms（GA）, the competitive ability of individuals and the algorithms ability of selecting elite individuals will get weak, and the evolution will stop when the colony grows mature. Therefore, the result is worthless repeated computation and long time waiting. In order to inspect the changing trend and improvement of the chromosomes in the evolution process and to enhance the search ability of the algorithms, we present an improved GA which adopts a method named off-line comparison and assistant selecting factor. Calculation experiments indicate that the off-line comparison method can efficiently control the computation progress and exactly look into the colony mature status and the evolution stop. Furthermore, the off-line method provides us a basis on which we decide when to improve the algorithms. Adapting the assistant selecting factor can keep the previous computing results and select those excellent individuals in the current colony. That the excellent individuals cross with the fresh individuals led into the colony makes the evolution go ahead continuously and consequently enhances greatly the probability that the improved GA succeeds in finding the qualified resolution.