Enabling the Design of Large-Scale Complex Engineered Systems using Self-Organizing Optimization Algorithms

由于这些系统对生产力和生活质量的影响，复杂工程系统的设计对社会至关重要。随着对性能和功能的期望不断提高，诸如电子设备、飞机、航天器或汽车等系统的复杂性已经大大增加。设计工程系统的过程本身已经变得复杂，到了这种复杂性变得难以管理的地步。该奖项支持通过使用优化算法来解决前所未有的复杂问题的新设计方法的研究。使用优化软件，汽车工程师已经可以通过改变汽车形状、尺寸和发动机特性来设计出最大限度地提高燃油经济性的汽车，同时满足所有安全约束。然而，对于涉及大量相互依赖的部件和参数的工程系统，现有的优化算法变得难以接受。该项目的目标是开发一套新的计算设计优化算法，能够以可扩展的方式优化大型复杂工程系统。这种可伸缩性将通过利用高性能并行计算资源和利用问题的数学结构来实现，并将其自动分解为可管理的组件。算法将建立在最近的数学发展，统一数值方法的解决和灵敏度分析的耦合系统。开发的算法将与美国国家航空航天局合作进行测试，以优化卫星的设计及其运行。最终的软件将是通用的，并将在开源许可协议下发布。

The design of complex engineered systems is of critical importance to society due to the impact these systems have on productivity and quality of life. The complexity of systems such as electronic appliances, aircraft, spacecraft, or cars has grown tremendously as the expectations for performance and functionality continue to increase. The process for designing engineered systems has itself become complex, to the point where this complexity is becoming unmanageable. This award supports research on new approaches for designing that can tackle problems of unprecedented complexity through the use of optimization algorithms. Using optimization software, an automotive engineer can already design a car that maximizes the fuel economy by varying the car shape, dimensions, and engine characteristics, while satisfying all safety constraints. However, for engineered systems that involve large numbers of components and parameters that depend on each other, existing optimization algorithms become unacceptably slow.The objective of this project is to develop a new set of computational design optimization algorithms that will be able to optimize large-scale complex engineered systems in a scalable way. This scalability will be achieved by taking advantage of high-performance parallel computing resources and by exploiting the mathematical structure of the problem, decomposing it automatically into manageable components. The algorithms will be built upon recent mathematical developments that unify numerical methods for the solution and sensitivity analysis of coupled systems. The developed algorithms will be tested in collaboration with the National Aeronautics and Space Administration for the design optimization of a satellite and its operation. The resulting software will be of general purpose, and will be published under an open source license agreement.