SINDRI: Synergistic utilisation of INformatics and Data centRic Integrity engineering

如果我们不更新、简化和自动化传统的手动设计、制造和终身评估流程，大型工业资产的长期、安全运营，包括关键的低碳能源发电基础设施，将变得令人望而却步。这一伙伴关系将开发一个全面的数字框架，包括一套模拟材料从进入使用到使用结束的行为的模型。该框架的开发将使其能够并入EDF的多物理数字双胞胎联合生态系统，取代目前的手动流程。该伙伴关系将：-使用最先进的表征工具，如亨利·罗伊斯研究所提供的工具，观察和量化各种传统和先进的制造工艺以及与发电有关的退化机制后材料的状况。先进的表征工具能够对大量材料进行评估，充分捕捉其内在的可变性和不均一性，这是直到最近还不可能的。-协调跨表征平台获得的数据结构，允许数据集成，并建立材料因制造和使用中退化而产生的行为的完整图景。-利用由艾伦·图灵研究所提供的投入开发的深度学习算法，询问表征工作获得的大型微观结构数据集。利用进入材料科学界的机会，这一伙伴关系将利用人类的专业知识来塑造深度学习算法，保持数据分析的保真度，同时消除缓慢的人类依赖方面。-开发和验证中尺度材料模型，作为材料行为的忠实数字孪生兄弟，模拟各种制造方法的进入使用条件，并在从特性分析中获得的知识的基础上，在服务中退化机制。-部署工业合作伙伴EDF开发的模型简化技术，以确定中尺度模型的主导参数。通过保留宏观尺度行为的控制参数，开发由中尺度行为提供信息但适用于组件尺度的工程模型。-根据宏观机械部件的高保真测试验证这些工程模型。-建立一个概率分析工具包，通过将概率理论应用于宏观机械工程模型的结果，并考虑与工厂运行范围相关的不确定性，来评估部件的安全置信度水平。-基于从EDF历史数据中提取的关键部件故障案例研究，对框架及其相关的模块材料模型进行最终验证和确认。通过在总体数字框架内以模块化方式开发各种材料行为模型，该伙伴关系将能够组装与历史案例研究相关的制造和退化模块，并使用概率工具包量化故障概率。-将框架及其模型作为组成材料数字孪生的基础纳入EDF的多物理数字孪生联合生态系统。

The long-term, safe operation of large industrial assets, including critical low-carbon energy generation infrastructure, will become prohibitively costly if we fail to update, streamline and automate traditional manual processes of design, fabrication, and life-time assessment. This partnership will develop an overarching digital framework encompassing a suite of models that simulate the behaviour of materials from their entry into service through to their end of life. The framework will be developed so it can be incorporated into EDF's federated ecosystem of multi-physics Digital Twins, replacing current manual processes. The partnership will:- Use state-of-the-art characterisation tools, such as those available at the Henry Royce Institute, to observe and quantify the conditions of the material after various conventional and advanced fabrication processes as well as degradation mechanisms relevant to power generation. Advanced characterisation tools enable the assessment of large volumes of materials that fully capture their inherent variability and inhomogeneity, impossible until very recently.- Harmonise the data structure obtained across characterisation platforms, allowing data integration and building a full picture of the material behaviour as a result of fabrication and in-service degradation.- Exploit deep learning algorithms, developed with input from the Alan Turing Institute, to interrogate large microstructural datasets obtained from characterisation work. Taking advantage of access to the materials science community, the partnership will shape the deep learning algorithms with human expertise, maintaining the fidelity of data analysis while removing the slow human-dependent aspects.- Develop and validate meso-scale material models as faithful digital twins of material behaviour, simulating the entry to service condition for various fabrication methods, and in-service degradation mechanisms, built on knowledge gained from characterisation analysis.- Deploy model reduction techniques developed by industrial partner, EDF, to identify the governing parameters of the meso-scale models. Develop engineering models that are informed by meso-scale behaviour but applicable to component- scale, by preserving the governing parameters of the macroscale behaviour. Validate these engineering models against high-fidelity tests on macro-mechanical components.- Build a probabilistic analysis toolkit that can assess the level of safety confidence of a component by applying applied probability theory to the results of the macro-mechanical engineering models informed by the material variability from meso-scale models and accounting for the uncertainty associated with the operational envelope of plants.- Undertake an ultimate verification and validation of the framework and its associated suite of modular material models based on extracted case studies of critical component failure from historic EDF data. By developing various models of materials behaviour in a modular way within the overarching digital framework, the partnership will be able to assemble the fabrication and degradation modules relevant to the historic case studies and quantify the probability of failure using the probabilistic toolkit.- Incorporate the framework and its models as the basis of a component material Digital Twin within EDF's federated ecosystem of multi-physics Digital Twins.