Constrained Minimax Optimization in Computer-Aided Design ofEngineering Systems

经典的非线性规划问题表述往往过于严格，无法充分描述工程设计问题。设计方法是基于具有三种类型规格的问题表述：硬约束、软约束和目标而开发的。与此方法相关，解决了两个核心问题：在工程环境中需要强大的优化算法，以及需要有效的方式向设计师传达有关可能的权衡设计的信息。算法开发的努力将集中在超线性收敛算法上。现有的方法将适应极大极小问题，并加以修改，以考虑工程设计环境中出现的各种要求。加强计算机和设计者之间沟通的工作将研究在不进行昂贵的模拟运行的情况下评估潜在权衡的方法。此外，将寻求有效的方式，以图形方式向设计师传达所获得的信息。最后，开发的各种技术将在DELIGHT系统中实施，并对各种类型的设计问题进行测试。

The classical nonlinear programming problem formulation is often too rigid to adequately describe an engineering design problem. A design methodology is developed based on a problem formulation featuring three types of specifications: hard constraints, soft constraints, and objectives. In relation with this methodology, two central issues are addressed: the need for powerful optimization algorithms in an engineering context and the need for efficient ways of conveying to the designer information about the possible tradeoff designs. The effort in algorithm development will focus on superlinearly convergent algorithms. Existing methods will be adapted to minimax problems and modified to take into account various requirements arising in an engineering design context. The work on enhancing the communication between computer and designer will investigate ways of evaluating potential tradeoffs without performing costly simulation runs. Also, efficient ways will be sought for graphically conveying to the designer the information obtained. Finally, the various techniques thus developed will be implemented in the DELIGHT system and tested on various types of design problems.