Self-learning universal AI to improve productivity in high-value metal additive manufacturing

金属基增材制造（AM）包括广泛的技术，如粉末床融合，粘合剂喷射和直接能量沉积。这些工艺主要用于航空航天、汽车、医疗和能源等要求苛刻的行业的制造。这些行业要求零件质量均匀，并具有与传统铸造或锻造工艺相同的性能。然而，金属增材制造极易出现部件故障-3D打印金属零件在打印过程中容易出现制造错误（例如，高孔隙率、变形、开裂）。高达17%的材料和宝贵的时间被浪费在失败的打印上，每年给金属制造业造成数百万英镑的损失。Matta正在开发一个自学习的通用AI数字平台，该平台利用最新的AI研究和云的力量为金属AM带来智能。机器学习用于检测印刷过程中的制造错误和效率低下，最终实现错误纠正和预防，以提高资源和能源效率。Matta的目标是在错误发生时自动检测错误，停止打印，并使用自动更正的设置重新启动打印。失败的打印是一种可以避免的资源，材料和能源浪费，最终是不可行的产品，可以浪费几天，如果不是几周的时间。在这个项目中，我们将把我们的技术应用于高价值的英国金属制造业。我们将收集金属增材制造过程的各种数据集，开发最先进的数据科学，构建自学习的通用AI模型，以自主检测错误，并将其应用于现场演示。该系统将被构建为通过云连接的金属打印机的分布式网络，因此每个金属打印机都可以从其他打印机的经验中学习。通过这种方式，将通过识别制造运行之间的相似性，创造最终实现自主错误预防所需的智能水平，从而大大提高该行业的生产力。该项目是开发和商业化全闭环增材制造系统的天然垫脚石，在该系统中，关键的制造参数可以在真实的时间内自动调整，无需人工干预-行业游戏规则改变者! Matta是剑桥大学的一名副教授。该项目对于我们从聚合物增材制造向高价值金属制造的延伸至关重要，而高价值金属制造的风险要高得多。自2021年成立以来，Matta还与制造技术中心、通用电气、剑桥和斯坦福大学等行业领导者建立了关系。

Metal-based additive manufacturing (AM) encompasses a wide range of technologies such as powder bed fusion, binder jetting and direct energy deposition. These processes are primarily used for manufacturing in demanding industries such as aerospace, automotive, medical, and energy. These industries require uniform part quality with properties like those achieved by conventional casting or forging processes.However, metal AM is highly prone to component failure - 3D printed metal parts are susceptible to manufacturing errors (e.g. high porosity, deformation, cracking) during the printing process. Up to 17% materials and valuable time is wasted via failed prints costing the metals manufacturing industry millions of pounds each year.Matta is developing a self-learning universal AI digital platform which leverages the latest AI research and the power of the cloud to bring intelligence to metal AM. Machine Learning is used to detect manufacturing errors and inefficiencies in the printing process, ultimately enabling error correction and prevention to improve resource and energy efficiency over time. Matta's aim is to autonomously detect errors as soon as they occur, stop the printing, and restart the print with automatically corrected settings. Failed prints are an avoidable waste of resources, materials, and energy for a product that is ultimately unviable, and can waste days if not weeks of time. In this project, we will apply our technology to the high-value UK metals manufacturing industry.We will collect diverse datasets of the metal AM process, develop state-of-the-art data science, build self-learning universal AI models to detect errors autonomously and apply them to a live demonstrator. The system will be built as a distributed network of metal printers connected via the cloud so each metal printer can learn from the experiences of others. In this way will create the level of intelligence required to ultimately achieve autonomous error prevention by recognising similarities between manufacturing runs, dramatically improving the productivity of the sector. This project is a natural stepping stone to the development and commercialisation of a total closed-loop AM system where crucial manufacturing parameters are automatically adjusted in real time without human intervention - an industry game changer!Matta is a spin-out from University of Cambridge. This project is critical to our extension from polymer AM to high-value metals manufacturing, where stakes are much higher. Since incorporating in 2021, Matta has also been building relationships with industry leaders such as the Manufacturing Technology Centre, GE, Cambridge, and Stanford.