In this paper, a new methodology was proposed for finding optimal robust-reliable parameter values of a space-based earth observation (SEO) mission based on a predefined multi-purpose platform considering influential uncertainties and human judgment. A deterministic optimization was performed using an evolutionary algorithm on the basis of platform capability and mission-required performance, simulated by a dynamic simulation-based model. A set of non-dominated solutions with different behaviors emerged. Then, samples of these solutions were selected and an uncertainty analysis was carried out. After that, a multiple attribute decision-making (MADM) problem was formed with two groups of attributes related to constraints violations and variation of objective function value having unknown weights. Dominant ranking of the non-dominated solutions was obtained by simulating this MADM problem for adequate times with different random weights (human judgment effects). In the next step, given the outputs obtained from solving MADM problem, an optimal robust-reliable solution could be determined utilizing two approaches, response surface methodology (RSM) and forming a new uncertainty-based multidisciplinary design optimization (UMDO) problem. The results showed that utilizing MADM approach not only brings the effects of human judgment into design problem but also accelerates convergence to optimal robust-reliable solution in multimodal problems by bounding the search space without any risk regarding getting stuck in local optimal regions.

中文翻译：

---

利用基于不确定性的 MADM 优化方法，通过考虑偶然的不确定性和人为判断的影响，为基于平台的产品寻找稳健可靠的设计参数

在本文中，基于预定义的多用途平台，考虑到有影响的不确定性和人类判断，提出了一种新的方法，用于寻找天基地球观测 (SEO) 任务的最佳稳健可靠参数值。在平台能力和任务所需性能的基础上，使用进化算法执行确定性优化，并通过基于动态仿真的模型进行模拟。出现了一组具有不同行为的非支配解决方案。然后，选择这些溶液的样品并进行不确定度分析。之后，形成了多属性决策（MADM）问题，其中包含两组与约束违反和目标函数值变化相关的未知权重的属性。通过使用不同的随机权重（人为判断效应）模拟这个 MADM 问题足够的时间来获得非支配解的支配排序。在下一步中，给定从解决 MADM 问题中获得的输出，可以使用两种方法确定最佳鲁棒可靠解决方案，即响应面方法 (RSM) 和形成新的基于不确定性的多学科设计优化 (UMDO) 问题。结果表明，利用 MADM 方法不仅将人类判断的影响带入设计问题，而且通过限制搜索空间来加速多模态问题中最优鲁棒可靠解决方案的收敛，而没有任何陷入局部最优区域的风险。