Multi-objective optimization and parameter uncertainty in the design of low frequency electromagnetic devices and systems

低频电磁系统（电动机和发电机、传感器、执行器等）对现代社会的运作至关重要。事实上，据估计，一个普通家庭拥有几百个这样的设备，分布在娱乐系统、计算设备、家用电器、汽车等。它们是能源（在制造和使用中）和原材料的主要用户，因此，为此类系统的设计人员提供基于计算机的仿真工具非常重要，这些工具既可以准确预测给定设备的性能，又可以搜索可以提供最佳性能的特定设备，同时提高能源效率并降低生产成本。例如，为了提供真实电动机的物理精确模型，需要考虑电磁，热，结构和机械效应之间的相互作用。为了提高效率，需要对所有可能的设计进行有效的搜索，并对可以做出的各种设计决策进行审查。为了降低建造成本，需要一个模型，该模型可以考虑到制造中的差异，如尺寸误差和材料特性的变化。这里提出的工作目标是创建允许实现这种模拟和设计过程的组件。正在开发的算法和软件架构将针对工程师的台式或笔记本电脑上日益流行的新计算机架构，即具有大内存和大磁盘的多核处理器。这个体系结构提供了一个新的模拟环境，在这个环境中，现有分析包中的许多假设可能不再成立。如果成功的话，这项研究的结果将把目前的设计时间从几天到几周的计算机工作缩短到几个小时，这意味着第一个物理原型将是唯一需要的。预计将培养至少4名博士和3名硕士生。

Low frequency electromagnetic systems (electric motors and generators, sensors, actuators, etc.) are vital to the functioning of modern society. In fact, it is estimated that an average family owns several hundred of these devices distributed between entertainment systems, computing devices, domestic appliances, automobiles, etc. They are a major users of energy (both in their manufacture and use) and raw materials and, as such, it is important that designers of such systems are provided with computer based simulation tools which can both accurately predict the performance of a given device and search for a particular device which can provide the best performance, while increasing the energy efficiency and reducing production costs. To provide a physically accurate model of a real electric motor, for example, requires that the interactions between electromagnetic, thermal, structural and mechanical effects are taken into account. To improve efficiency requires an effective search of all the possible designs and an examination of the various design decisions which can be made. To reduce the cost of construction requires a model which can take into account variances in manufacturing such as errors in dimensions and changes in material properties. The goal of the work being proposed here is to create the components which will allow such a simulation and design process to be implemented. The algorithms and software architectures being developed will be targeted at the new computer architectures which are becoming prevalent on an engineer's desktop or laptop computer, i.e. a multi-core processor with a large memory and massive disks. This architecture provides a new simulation environment where many of the assumptions made in existing analysis packages may no longer hold. If successful, the results of the research being proposed will reduce current design times, which run into days and weeks of computer effort, to a few hours and will mean that the first physical prototype built will be the only one needed. It is expected that the work funded by this grant will result in the training of at least four Ph.D. and three Master's students.