The Collaborative Optimization (CO) is an easily accepted technique which has been utilized by engineering designer. There are three problems on how to actualize CO in iSIGHT; (1) the ξ called the relaxation parameter has some subjectivity and blindness in choosing an exact value; (2) CO will be converged sometimes at a point which does not meet the constraint of sub-discipline, if the actual optimal point is at the boundary of the sub-discipline constraint; (3) a technique called the Dynamic Relaxation could not improve the ability of universal search under some conditions. We emphasize that, in the full paper, we explain the methods of solving these three problems in some detail. The core of our explanation consists of: "The first topic is how to confirm an exact ξ. In the first topic, the optimal points of CO based on the diverse ξ have been calculated, and an advisable region of ξ has been given. The second topic is the case that the optimal point violates the constraint of sub-discipline. A method called the Shrunk Region improves this situation based on the confirmed ξ. The third topic is about the Dynamic Relaxation. In this topic, an improved Dynamic Relaxation has been developed. Section 1 describes the flow of CO by iSIGHT. The history plot of response/is shown by Fig. 5. Subsections 2.1, 2.2 and 2.3 introduce how the problems have been discovered, analyzed and solved by certain methods. Tab. 1 gives the optimal points of CO based on the diverse ξ. Fig. 6 shows how the Shrunk Region improves this situation based on the confirmed ξ and enhances its applicability. An improved Dynamic Relaxation has been introduced in subsection 2.3 which increases the probability of finding global optimal point, and the history plot of response/is shown by Fig. 10.".%协同优化算法的学科自治,系统协调思想易被工程设计者所接受,有很广的发展前景.文中基于iSIGHT软件,通过对该算法的研究,发现存在3个问题:松弛因子ξ的取值带有一定的主观性和盲目性；系统最优点若在约束边界上,协同优化算法往往会收敛到一个不满足子学科约束的最优值；动态松弛法在某些情况下有可能无法达到预期的增强全局搜索能力的目的.针对上述问题提出如下解决方法:结合算法的准确性和快速性给出了ξ的推荐取值区间；采用区间微缩法消除了系统最优值违背子学科约束的情况,提高了算法的适用性；改进了动态松弛法使其全局搜索能力进一步增强.最后经算例验证了上述方法的有效性,在一定程度上充实了协同优化算法体系.
展开▼
