Developing Key Technologies towards an Engineer Centered Quantitative Design Methodology

工程设计作为一个创新的、复杂的、高度受限的过程，需要实用的方法和工具。关于设计自动化的研究已经有几十年的时间了，然而人类还远远没有能够“自动化”工程设计。 这项工作是基于这样一种信念，即人类，工程师，应该保持在创造力和创新的设计活动的中心。然而，而不是依赖于个人经验，工程师应该支持定量设计方法，为他们提供高层次的设计信息，以帮助他们做出设计决策。本研究旨在开发实现这种定量设计方法的关键技术。**工程师需要定量支持的设计活动是什么？ 工程师经常探索具有可互换设计目标和约束的不同设计方案。此外，工程师需要回答“假设”问题，找到敏感的设计参数，确定不同场景的最佳设计，并向管理层解释最佳设计。 这些问题的答案只能通过定量分析、整合和优化来充分获得。 为了满足工程师的需求，确定了三个关键技术，并将在本研究中发展为以工程师为中心的定量设计方法：1）智能优化管理; 2）以工程师为中心的设计交互;和3）分布式工程计算环境。 智能优化管理技术有望将工程师从优化专家中解放出来。为了实现这一目标，我们建议使用一个统一的优化配方，工程师可以自由地指定设计目标和约束条件，探索各种设计方案。 其次，我们将使用数据挖掘技术来获得有关设计问题的知识，这些知识可以用来辅助优化。 第三，我们的目标是开发一种优化策略，其搜索方案和参数通过反馈控制回路自动调整。 所有这些方法都将建立在我们实验室在设计优化方面的优势之上，并将实现自动优化过程，而无需工程师手动选择优化算法或其参数。对于以工程师为中心的设计交互，我们建议使用代理人进行快速可视化，应用数据挖掘技术进行预处理和后处理，并开发先进的视觉技术来支持各种设计研究。 为了开发分布式工程计算环境，我们将定制一个典型的分布式计算系统，以满足工程设计的独特需求。新的配置和负载平衡方案将被设计。** 三十年来，设计自动化研究社区一直专注于基于纯数学的设计自动化方法。所提出的研究革命性的传统设计自动化的研究，使工程师的设计中心。所提出的方法也不同于缺乏坚实的定量基础的定性设计方法。这种方法将使工程师能够系统有效地探索和搜索最佳设计，并利用现代工程分析的强大功能。通过提出的定量设计方法，工程师将做出明智的决策，并期望更有效地产生更好的产品和新工艺。因此，拟议的研究对加拿大制造业的影响是重大的，并将通过与行业伙伴的密切合作来显示实际利益和及时的技术转让。

Engineering design, as an innovative, complex, and highly constrained process, demands practical methodologies and tools. There have been a few decades of research on design automation, yet human beings are far from being able to "automate" engineering design. This work is based on the belief that human beings, i.e., engineers, should remain at the center of the design activities for creativity and innovation. However, instead of relying on personal experience, engineers should be supported with a quantitative design methodology, which provides them with high-level design information to assist in making design decisions. This research aims at developing key technologies towards such a quantitative design methodology.**What are the design activities that an engineer needs quantitative support for? An engineer often explores different design scenarios with interchangeable design objectives and constraints. Also, an engineer needs to answer "what-if" questions, find sensitive design parameters, identify the optimal design for different scenarios, and explain the optimal design to the management. Answers to these questions can only be adequately obtained through quantitative analysis, integration, and optimization. To address engineer's needs, three key technologies are identified and will be developed in this research towards an engineer-centered quantitative design methodology: 1) intelligent optimization management; 2) engineer-centered design interaction; and 3) distributed engineering computing environment. The intelligent optimization management technology is expected to free-up engineers from being experts in optimization. To achieve this, we propose to use a unified optimization formulation, with which engineers can freely designate design objectives and constraints to explore various design scenarios. Secondly, we are to use data mining technologies to gain knowledge about the design problem, which can be utilized to assist the optimization. Thirdly, we aim to develop an optimization strategy whose search schemes and parameters are automatically tuned through a feedback control loop. All of these methods are to be built on the strength of our lab in design optimization, and will enable an automatic optimization process without asking engineers to manually select optimization algorithms or their parameters. For engineer-centered design interaction, we propose to use surrogates for fast visualization, apply data mining techniques for pre- and post-processing, and develop advanced visual techniques to support various design studies. To develop a distributed engineering computing environment, we will tailor a typical distributed computing system to satisfy unique requirements of engineering design. New configuration and load balancing schemes are to be designed. **For three decades, the design automation research community has been focusing on pure mathematics based design automation methodologies. The proposed research revolutionizes the conventional design automation research by bringing engineers to the center of the design. The proposed methodology also differs from qualitative design methodologies that lack solid quantitative foundation. This methodology will enable engineers to systematically and efficiently explore and search for the optimal design and leverage strong capabilities of modern engineering analyses. With the proposed quantitative design methodology, engineers will make informed decisions, and better products and novel processes are expected to be generated more efficiently. The impact of the proposed research to the manufacturing sector of Canada is therefore significant and will be shown through close collaborations with industry partners for practical benefits and timely technology transfers.