机翼结构优化设计的多目标分布估计算法  

MULTI-OBJECTIVE ESTIMATION OF DISTRIBUTION ALGORITHM FOR OPTIMIZATION DESIGN OF WING STRUCTURE

作　　者：钟小平[1] 李为吉[1] 苟仲秋[1] 安伟刚[1] 

机构地区：[1]西北工业大学航空学院,西安710072

出　　处：《机械强度》2008年第4期601-605,共5页Journal of Mechanical Strength

摘　　要：提出一种采用贝叶斯网络表示概率模型的多目标分布估计算法(multi-objective estimation of distribution algo-rithm,MEDA)。通过构建这样的网络模型,对模型进行抽样生成新个体,再对新个体进行变异操作,以增加种群的多样性,提高算法的搜索能力。这种生成个体的方法结合非劣分层以及截断选择机制,可以获得很好地逼近多目标问题的Pareto前沿且分布均匀的非劣解集。用MEDA对某高空长航时无人机机翼结构进行多目标优化设计,找到高质量的非劣解集,为设计者作决策提供很好的参考依据。根据所得到的非劣解集,设计者可以很好地进行权衡折衷,找出最符合要求的设计方案。同时,还可以了解各目标之间的变化关系,定量化了解一个目标的改进将导致其余目标恶化的程度。研究表明,多目标分布估计算法可以有效求解复杂结构的优化设计问题。A multi-objective estimation of distribution algorithm （MEDA） is proposed. This algorithm uses decision-tree-based Bayesian network as probabilistic model to encode conditional dependencies among variables. By building and sampling the probabilistic models, the algorithm reproduces the genetic information of the next generation. Combined such a reproduction mechanism with the non- dominated sorting and truncated selection techniques, MEDA can approximate the probability density of solutions lying on the Pareto front. In MEDA, polynomial mutation operator is incorporated in order to enhance exploration and maintain diversities in the populations. The proposed algorithm is used to optimize the wing structure of a high altitude long endurance unmanned aviation vehicle （UAV）. And a well-spread nondominated solution set is obtained. With the obtained set, the designer can select the most desirable design by making tradeoff among the criteria. Further more, the designer can understand how much the improvement in one objective can lead to degradation of other objectives. The results indicate that this new algorithm is an effective multi-objective optimization algorithm and can be applied to solving complex engineering optimization problems.

关 键 词：机翼结构优化 无人机 多目标分布估计算法 贝叶斯网络 

分 类 号：O224[理学—运筹学与控制论] V224[理学—数学]