A NEW FRAMEWORK FOR HYBRID THROUGH-PROCESS MODELLING, PROCESS SIMULATION AND OPTIMISATION IN THE METALS INDUSTRY

IMMPETUS（微观结构和机械过程工程研究所：谢菲尔德大学）成立于1997年，致力于跨系统，机械和冶金工程学科进行真正集成的跨学科研究，解决金属加工行业的关键问题。在过去的十年里，IMMPETUS独特的跨学科研究已经获得了国内和国际的赞誉，其系统驱动的方法来优化各种金属工艺路线的工艺和性能。利用系统工程，我们针对和优化实验，在知识不完整的特定领域开发基本的物理冶金学，为模型推导、测试和验证提供信息。对于我们研究的复杂工业过程，没有足够的基础知识来构建真正的基于物理的模型。为了涵盖现有基于物理的模型无法充分描述的棘手因素，我们使用混合模型，以独特的方式合并离散数据，基于知识和基于物理的模型，以提供前所未有的预测模型能力的精度。所有的建模是通过使用世界一流的实验技术阵列进行验证。该提案包括与我们的工业合作伙伴一起建造的12个项目，以回答以下问题：1。我们如何为“全过程”建模制定“通用”框架，以实现金属的“首次正确”生产？2.哪些冶金和热机械变量影响金属的微观结构，从而影响金属的最终性能，但现有模型尚未完全描述？3.因果关系（确定性行为）以及不确定性（异质性，随机行为）如何影响加工过程并影响金属的最终性能？4.最适合回答上述1、2和3的具体建模策略是什么？5.使用第4章中的模型，我们能否确定给定工艺路线的可实现属性，以及如果某个特定属性无法实现该怎么办？6.使用5，我们如何优化工艺路线？工作方案分为四个主题，所有这些主题都是相互依存和相互交织的。物理系统将旨在发展基本的物理冶金知识是不够的，在包括微观结构的不均匀性，以及动态和非线性的工艺条件。在建模系统中，物理冶金，机械工程和系统工程将完全集成，通过开发新的建模方法，以及现有的最先进的建模本身产生新的方法的耦合。工艺模拟将涉及扩大重点实验室实验的规模，以准确和完整地模拟相关的工业工艺路线，并通过适当的工厂试验对其进行验证。系统优化将作为整合这些主题的强大工具，通过仔细调整模型结构/参数，将成为我们向特定工业赞助商和更广泛学术界转让技术的核心。

IMMPETUS (Institute for Microstructural and Mechanical Process Engineering: The University of Sheffield) was founded in 1997 to undertake truly integrated interdisciplinary research across the disciplines of systems, mechanical and metallurgical engineering, addressing key issues in the metals processing industry. Over the last ten years the unique inter-disciplinary research produced by IMMPETUS has secured national and international acclaim for its systems driven approach to process and property optimisation of a wide range of metals process routes. Using systems engineering we target and optimise experiments to develop basic physical metallurgy in specific areas where knowledge is incomplete, to inform model elicitation, testing and validation. For the complex industrial processes we investigate, there is insufficient basic knowledge to construct true through-process physically based models. In order to cover the intractable factors not adequately described by the existing physically based models, we use hybrid models that merge discrete data, knowledge-based and physically-based models in a unique manner to give unprecedented precision in predictive model capability. All the modelling is verified through the use of a world class array of experimental techniques. The proposal comprises 12 projects which have been constructed in conjunction with our industrial collaborators in order to answer the following questions: 1. How do we formulate a 'generic' framework for 'through-process' modelling to achieve 'right first-time' production of metals?2. Which of the metallurgical and thermomechanical variables affect the microstructure and therefore the final properties of metals, but are not yet fully described by existing models?3. How do causalities (deterministic behaviours) as well as uncertainties (heterogeneities, random behaviours) influence the processing and affect the final properties of metals?4. What are the specific modelling strategies 'best' suited for answering 1, 2, and 3 above?5. Using the elicited models in 4, can we identify the achievable properties for a given process route, and what to do if a particular property is not achievable?6. Using 5, how do we optimise the process route?The programme of work is presented as four themes, all of which are inter-dependent and interwoven. PHYSICAL SYSTEMS will be aimed at developing basic physical metallurgical understanding where knowledge is inadequate, in areas including microstructural heterogeneities, and process conditions that are dynamic and non-linear. In MODELLING SYSTEMS, the physical metallurgy, mechanical engineering and systems engineering will be fully integrated, both through the development of new modelling approaches, and the coupling of existing state-of-the-art modelling that in itself produces new methodologies. PROCESS SIMULATION will involve the upscaling of focused laboratory experiments to accurately and completely simulate the relevant industrial process routes and validate them through appropriate mill trials. SYSTEMS OPTIMISATION will act as a powerful vehicle for integrating these themes and via a careful tuning of model structures/parameters will be core to our technology transfer to our will target specific industrial sponsors and to the wider academic community.