船舶强框架序贯代理模型辅助遗传优化方法  

Genetic optimization method of ship strong frame based on sequential surrogate model

作　　者：汪俊泽 王元 易家祥 韩涛 江璞玉 吴嘉蒙[2] 程远胜[1] 刘均[1] WANG Junze;WANG Yuan;YI Jiaxiang;HAN Tao;JIANG Puyu;WU Jiameng;CHENG Yuansheng;LIU Jun(School of Naval Architecture and Ocean Engineering,Huazhong University of Science and Technology,Wuhan 430074,China;Marine Design and Research Institute of China,Shanghai 200011,China)

机构地区：[1]华中科技大学船舶与海洋工程学院,武汉430074 [2]中国船舶及海洋工程设计研究院,上海200011

出　　处：《中国舰船研究》2021年第4期44-52,共9页Chinese Journal of Ship Research

摘　　要：[目的]共同结构规范(CSR)要求下的船舶强框架结构优化存在约束条件多、计算耗时、可行性判断复杂的特点。应用静态代理模型辅助优化算法求解该问题时,因其关注的是模型的整体预测精度,故在样本容量较小的情况下无法保证关键区域的模型预测精度。针对该问题,提出基于序贯代理模型辅助遗传算法的强框架优化方法。[方法]首先,分析CSR对强框架结构的约束要求,根据约束类型,将原始的675条约束缩减为2条积极约束,再对目标函数和约束函数建立代理模型。然后,基于可行性准则,利用遗传算法对代理模型进行优化求解,得到优化解后,计算优化解的真实响应并更新代理模型,再利用期望可行性函数(EFF)准则更新约束代理模型,提高代理模型在约束边界上的精度,如此迭代求解多次,最终得到满足约束条件的全局最优解。[结果]强框架优化结果显示,所提序贯代理模型算法能够在相同,甚至更少的计算资源下得到优于基于静态代理模型优化算法的优化解,最终实现设计区域减重达15.55%。[结论]提出的序贯代理模型算法显著优于静态代理模型算法,在复杂约束下的船舶结构优化问题上有着较好的应用价值。[Objectives]The optimization design of a ship strong frame structure under the requirements of the common structural rules(CSR)is a complex and time-consuming problem.Moreover,its tremendous constraints make it difficult to judge the feasibility of any design scheme.As the approach aims at global accuracy,when adopting the static surrogate-assisted evolutionary algorithm to solve this problem,the prediction of key areas will be distorted in the case of small size samples.Aiming at the above problem,a strong ship frame optimization method based on a sequential surrogate-assisted genetic algorithm is proposed.[Methods]First,the constraints of a strong frame structure based on CSR are analyzed,and all 675 constraints are reduced to 2 positive constraints according to constraint type.Then,surrogates for objective functions and constraint functions are constructed,and a genetic algorithm based on the feasibility principle is adopted to find the optimized solution.The true response of the solution is then calculated and the surrogates updated.In addition,the expected feasibility function(EFF)criterion is applied to update the constraint surrogates in order to refine the prediction accuracy at the constraint boundaries.The above procedures are iterated several times,and the optimized global feasible solution is finally obtained.[Results]The proposed method can obtain a better solution than the static surrogate-based algorithm with a lower computational burden,and the weight of the design area is finally reduced by 15.55%.[Conclusions]The proposed sequential surrogate-based algorithm is superior to the static surrogate-based algorithm,and possesses good application value in the optimization of ship strong frame structures under complex constraints.

关 键 词：强框架 结构优化 序贯代理模型 期望可行性函数 遗传算法 共同结构规范 

分 类 号：U663.2[交通运输工程—船舶及航道工程]