Artificial Intelligence X-ray Imaging for Sustainable Metal Manufacturing (AIXISuMM)

金属制造业占全球二氧化碳排放量的8%，如果要在2050年前实现碳中和，我们迫切需要向更可持续的过程过渡。在这个项目中，我们解决了基础科学和理解，以允许更高地利用低嵌入碳，高杂质回收金属作为金属制造的原料。目前的制造方法高度依赖于能源密集型的原生金属，因为它们依赖于严格控制的成分，杂质含量非常低，以提供所需的材料性能。我们相信，利用多模态x射线成像和在线人工智能的综合能力，可以提供变革性和高效的方法，以使用更高比例的低嵌入碳回收材料作为原料来制造高品位金属合金。我们将开发一种新的整体表征系统，包括新开发的硬件和名为人工智能x射线成像（AIXI）的人工智能算法，作为一种智能工具，用于研究在实验条件下富含杂质的合金的凝固情况，可与工业工艺（如连铸、直接冷铸、形状铸造和增材制造）中发现的情况相比较，适用于广泛的铝和钢合金成分。AIXI将提供比现有方法显著的优势，因为AI将嵌入到数据采集系统中，用于实时解释原始数据，大大降低了数据分析所需的复杂性和时间，并显着提高了系统的分析能力。新的知识将使我们最终了解杂质和少量合金添加物在发展中的凝固组织中所起的作用，并开发方法来减轻它们的有害影响。它还将促进向更全面的合金设计方法的转变，在这种方法中，凝固微观结构的设计既能提供增强的性能，又能在对环境影响最小的情况下促进后续的下游工艺。新获得的知识将促进“可持续”合金科学的发展，这将：通过减少用能源密集型原生金属稀释回收废料的需求，提高金属的可回收性；鼓励更多地使用低品位废料，这些废料在英国广泛存在，但目前出口；减少下游加工步骤的数量（工艺强化），特别是热处理操作；通过减少对严格成分规范的依赖，简化组件的可恢复性；并通过改善对最终微观结构的控制来提高材料的性能。我们将发现并应用缺失的科学来控制相变，以创造更良性和更耐杂质的微观结构，并允许更有效地使用昂贵和潜在稀缺的合金添加剂，这将大大减少二氧化碳密集型金属工业的资源使用。此外，我们设想开发的硬件/人工智能分析的应用可能会促进材料科学等许多领域的快速科学发展，在这些领域，高效，快速收集和分析复杂和大型多模态数据集对于解锁必要的理解至关重要

Metal manufacturing is responsible for 8% of global CO2 emissions and if carbon neutrality is to be achieved by 2050, we critically need to transition to more sustainable processes. In this project we address the underlying science and understanding to allow a higher utilisation of low embedded-carbon, higher impurity recycled metal as a feedstock for metal manufacturing.Current manufacturing approaches are highly dependent on energy-intensive primary metal as they rely on tightly controlled compositions with very low impurity contents to provide the required materials properties. We believe that the new understanding needed to provide transformative and efficient methods to manufacture high grade metal alloys using a much higher fraction of lower embedded-carbon recycled material as a feedstock can be delivered by leveraging the combined power of multi-modal X-ray imaging and in-line artificial intelligence.We will develop a new wholistic characterisation system comprising both newly developed hardware and AI algorithms named Artificial Intelligence X-ray Imaging (AIXI) as an intelligent tool to investigate the solidification of impurity-rich alloys in experimental conditions comparable to those found in industrial processes such as continuous casting, direct chill casting, shape casting and additive manufacturing for a wide range of aluminium and steel alloy compositions. AIXI will provide a significant advantage over existing approaches as AI will be embedded in the data acquisition system and used to interpret raw data in real-time, drastically reducing the complexity and time required for data analysis and significantly increasing the analytical power of the system. The new knowledge will allow us to finally understand the role that impurities and minor alloy additions play in the developing solidification microstructure, and to develop methodologies to mitigate their deleterious effects. It will also promote a shift to a more holistic approach for alloy design in which the solidification microstructure is engineered to both provide enhanced properties and to facilitate subsequent downstream processes with minimised environmental impact.The newly acquired knowledge will foster the development of science for `sustainable' alloys, which will: enhance metal recyclability by reducing the need for dilution of recycled scrap with energy intensive primary metal; encourage greater use of lower-grade scrap, widely available in the UK but currently exported; decrease the number of downstream processing steps (process intensification), especially heat treatment practices; simplify component recoverability by reducing the reliance on tight compositions specifications; and enhance materials properties by improving control over the final microstructure. We will uncover and apply the missing science to control phase transformations to create more benign and impurity tolerant microstructures and allow more efficient use of expensive and potentially scarce alloy additions, which will substantially cut resource use in the CO2-intensive metal industries. Furthermore, we envisage that the application of the developed hardware/AI analysis could potentially facilitate rapid scientific development in many fields of materials science and beyond where efficient, rapid collection and analysis of complex and large multi-modal datasets is critical to unlock the necessary understanding