Data-driven, Reliable, and Effective Additive Manufacturing using multi-BEAM technologies (DREAM BEAM)

激光粉末床融合（LPBF）增材制造（AM）通过将材料一层一层地连接在一起，将数字设计转化为功能产品。它提供灵活，可持续的制造能力和短的产品开发时间，为全球业务生产具有复杂几何形状的高价值组件，包括航空航天，汽车和生物医学领域。全球增材制造市场预计将从60亿美元（2016年）增长到260亿美元（2022年），从而在全球范围内推出了发展增材制造技术的重大举措，包括“英国工业战略”、“弗劳恩霍夫增材制造联盟”、“中国制造2025”和“美国制造”。尽管增材制造具有关键优势，但行业正面临着将增材制造技术用于安全关键产品的技术挑战，例如螺旋桨和涡轮叶片等。这些产品可能由于加工缺陷的存在而表现出较差的机械性能。为了生产高性能的增材制造产品，利益相关者必须了解增材制造过程中的过程和缺陷动力学，然而，由于在毫秒内发生的快速，复杂的激光物质和多相（固体-液体-气体-等离子体）相互作用，它们很难表征。该项目涉及伦敦大学学院和世界领先的增材制造工业合作伙伴（雷尼绍有限公司）、激光技术（STFC -中央激光设备）、机器学习（STFC -科学机器学习小组）、超快速成像（欧洲同步辐射设备）和过程模拟（欧洲航天局），共同开发工程解决方案，以理解、评估和控制增材制造中的过程-结构-性能-性能关系。该项目预计将收集广泛的数字数据，可用于开发数据驱动，可靠和高效的增材制造过程。首先，将开发并部署一种独特的化学成像工具，以监测和评估LPBF过程中的金属汽化过程，时间分辨率为200 kHz。这些结果将通过旗舰级超快速x射线成像实验进行交叉验证，该实验使用户能够以微米分辨率和高达1 MHz的时间分辨率看到熔体池内部和LPBF过程中的缺陷动态。AM的相关化学和x射线成像将成为研究AM动态行为和多相相互作用的一种改变游戏规则的表征技术。它将带来新的理解，在制造过程中引入的缺陷，并提出改进整个过程的方法。其次，我们将推动新型光束整形技术的发展，以控制熔合过程中的热输入，最大限度地减少金属汽化和LPBF过程中的缺陷形成。波束整形技术的性能将通过相关成像进行评估和验证。第三，通过该项目收集的所有数字数据将用于构建、训练和部署用于过程控制的机器学习（ML）模型，即ML引导的过程控制。它们还将用于验证、验证和推进开源高保真过程仿真模型，该模型分析增材制造中的多相和多物理场相互作用，可扩展到其他先进制造工艺。除了发展新技术外，该计划亦会为初入职的研究人员提供机会，向公众、业界和科学界传播他们的研究成果，促进知识交流和技术转移活动。

Laser powder bed fusion (LPBF) additive manufacturing (AM) transforms digital designs into functional products by joining materials together, layer upon layer. It offers flexible, sustainable manufacturability and short product development time to produce high-value components with complex geometries for business across the globe, including aerospace, automotive, and biomedical sectors. The global market for AM is expected to grow from $6b (2016) to $26b (2022), resulting in major initiatives launched across the globe to grow AM technologies, including "UK Industrial strategy", "Fraunhofer Additive Manufacturing Alliance", "Made in China 2025", and "America Makes". Despite the key advantages of AM, industries are facing technical challenges to use AM technology for safety-critical products, e.g. propellers and turbine blades, etc. These products may exhibit poor mechanical performance due to the presence of processing defects. To produce high-performance AM products, the stakeholders must understand the process and defect dynamics during AM, however, they are difficult to characterise due to the fast, complex laser-matter and multi-phase (solid-liquid-gas-plasma) interactions which occur in milliseconds. This project involves UCL and world-leading industrial partners in AM (Renishaw plc.), laser technologies (STFC - Central laser facility), machine learning (STFC - Scientific Machine-learning group), ultra-fast imaging (European Synchrotron Radiation Facility) and process simulations (European Space Agency) to co-develop engineering solutions to understand, evaluate, and control the process-structure-property-performance relationships in AM. This project is expected to collect a wide range of digital data that can be used to develop a data-driven, reliable and efficient AM process. Firstly, a unique chemical imaging tool will be developed and deployed to monitor and evaluate the metal vapourisation process during LPBF with a temporal resolution of 200 kHz. These results will be cross-validated by flagship ultra-fast X-ray imaging experiments which enable users to see inside the melt pool and defect dynamics during LPBF at micron resolution and a time resolution of up to 1 MHz. Correlative chemical and X-ray imaging of AM will be a game-changer characterisation technique to study the dynamic behaviour and multiphase interaction in AM. It will bring new understanding by which defects are introduced during AM and suggest ways to improve the overall process. Secondly, we will make advancement of novel beam shaping technologies to control the heat input to the fusion process, minimising metal vapourisation and defect formation during LPBF. The performance of the beam-shaping technologies will be assessed and verified by correlative imaging. Thirdly, all the digital data collected through this project will be used to build, train and deploy machine learning (ML) model(s) for process control, i.e. ML-guided process control. They will also be used to verify, validate, and advance an open-source high fidelity process simulation model that analyses multi-phase and multi-physics interactions in AM, which can be extended to other advanced manufacturing processes. Besides the development of new technologies, this project will also provide opportunities for early-career researchers to disseminate their research to the public, industries, and scientific communities, promote knowledge exchange and technology transfer activities.