改进的贝叶斯优化算法及应用

提出了一种改进的贝叶斯优化算法。该算法通过引入免疫算法中的亲和度和浓度概念,将个体适应度概率和个体浓度概率相结合,形成贝叶斯优化算法选择优良个体的依据。这样,由低浓度、高适应度个体组成优良个体种群,能够保持种群的多样性,提高算法的性能。本文利用改进的贝叶斯优化方法对十杆平面桁架结构、二十五空间桁架结构进行优化设计,取得了满意的结果。

A Modified Bayesian Optimization Algorithm

The probabilities of Bayesian optimization algorithm(BOA) presented in the paper incorporate the concepts of affinity and thickness in immune algorithm to form selectional criteria for promising individuals.The selectional criteria are combinations of probabilities of individual fitness and individual thickness,and can select high-fitness and small-thickness individuals.Thus the algorithm can maintain diversity in selecting promising individuals to enhance the performance of the BOA.The modified BOA is applied to the design optimization of ten-bar trusses and twenty-five-bar trusses,the results obtained from which are compared with those from the BOA.The comparison shows that the modified BOA can accomplish design optimization more reliably and effectively.