Real-time defect detection is critical for the large-scale adoption of additive manufacturing in aerospace and other demanding applications. Long Beach, California-based Relativity Space has been awarded an $8.7 million contract with the U.S. Air Force Research Laboratory (AFRL) to explore real-time defect detection in additive manufacturing.
3D printed rockets
© Relativity Space
A true digital thread
The two-year research contract comes from the AFRL Materials and Manufacturing Directorate at Ohio Air Force Base in the United States. The project will be completed using the Stargate 3D AM platform at Relativity's facility in Long Beach, California.
According to 3D Science Valley, Tim Ellis, the founder of Relativity Space, has a deep insight that he thinks that what is not really understood about 3D printing in the market is that the disruption of 3D printing to manufacturing is actually more like the transition from gas internal combustion engines to electric, or from on-premise services to the cloud, 3D printing is a cool technology, but more importantly, 3D printing is actually software and data-driven manufacturing and automation technology.
According to Relativity Space's patent, for the directed energy deposition 3D printing process, process error correction can be performed by artificial intelligence and process simulation data can be provided by performing finite element analysis (FEA), finite volume analysis (FVA), finite difference analysis (FDA), computational fluid dynamics (CFD) calculations, or any combination thereof.
This is an even more milestone for Relativity Space, where AFRL will work with Relativity Space to conduct real-time defect detection for large-scale additive manufacturing, with the goal of exploring various on-site process monitoring and building non-destructive evaluation technologies. In the National Defense Authorization Act, the U.S. Congress directed the U.S. Department of Defense to study how additive manufacturing could be used to accelerate the production of aerospace components and vehicles. It is in this context that the project was created, in addition, the legislator asked the Ministry of Defense to create a network of domestic suppliers to help evaluate these technologies.
Relativity Space will develop and validate a real-time defect detection system that will detect, locate, and classify defect types during the 3D printing process, and then this data will be aggregated to enable a true digital thread.
Artificial intelligence is used for process control
© 3D Science Valley White Paper
artificial intelligence
According to 3D Science Valley's market understanding, the use of ML machine learning for real-time and near-real-time diagnostic predictions of AM additive manufacturing processes is multifaceted:
1. ML1 - Parameter Setting - Machine learning-based optimization algorithms provide the best possible values of process parameters that define the optimal machining route that is most likely to avoid defects, and in-situ process optimization strategies using reinforcement learning (RL) algorithms have been demonstrated.
2. ML2-Diagnostics – Continuously evaluate performance as required: Machine learning-based defect, anomaly, and error detection algorithms can be used for diagnostic capabilities to detect defects, anomalies, and errors (i.e., process deviations) in the production process. By comparing the process to performance requirements, building processes from sensory data, these algorithms need to run online to provide instant warnings based on rapid analysis of sensor signals in the field.
3. ML23 - Prediction – Continuously Adjust Machining Results by Predicting and Making Control Decisions: Machine learning-based predictive algorithms allow predictive process control during construction to ensure that the additive manufacturing process remains within specification to meet performance requirements, for example, avoiding, mitigating, or repairing defects, or minimizing process deviations, these algorithms must also work online.
Smarter brains
Globally, additive manufacturing is on the eve of artificial intelligence giving it wings. According to the German ACAM Aachen Additive Manufacturing Center, 3D printing companies in the world generally do not achieve good profitability, a key point is from the perspective of application industrialization, the manufacturing model that can achieve profitability should be the economic benefit of the digital driven end-to-end manufacturing process chain as the core, and the current 3D printing is caught in a dilemma, often when the scale is expanded, the ensuing production cost increases in stages, which in turn makes it very challenging to achieve profitability。 Additive manufacturing will develop in the direction of software and data-driven self-evolving intelligent manufacturing technology, and the application of artificial intelligence will make hardware have more "smart brains", more "sensitive nerves" and "more accurate hands", making processing more efficient.
According to the Future of AI in Progression, only when the industry foresees that some high-value applications can be implemented, will the related expensive technologies have the opportunity to continue to develop and mature. If a technology solves a particularly critical need, some companies are often willing to pay for the huge investment or even loss in the early stage of the development of the technology, in exchange for the possibility of relying on the technology to expand and obtain higher profits later. It can be seen that at the historical node where artificial intelligence has become the intelligent hub of intelligent manufacturing, scientific research institutions will continue to make great progress in promoting 3D printing to become a software and data-driven self-evolving intelligent manufacturing technology, and the future belongs to those enterprises that can see the trend and integrate the trend into their own development.
Relativity Space is one of those companies that sees trends and incorporates them into their own development.
Better preparation
根据3D科学谷的市场观察，早在2021年，Relativity Space专门申请了使用机器学习对增材制造过程进行实时自适应控制Real-time adaptive control of additive manufacturing processes using machine learning的专利。
Relativity Space's patented methods and systems for automated object defect classification and adaptive real-time control can provide rapid optimization and adjustment of process control parameters used in response to changes in process or environmental parameters to improve process stability, increase throughput, and improve the quality of parts produced. These methods and systems are suitable for part manufacturing in a variety of different technical fields and industries, including the automotive industry, the aerospace industry, the medical device industry, the consumer electronics industry, and more.
According to the market judgment of 3D Science Valley, in the near future, the next step of artificial intelligence will be to cross a single 3D printing equipment to achieve coordination and process optimization between equipment. With the help of software, the young 3D printing industry is looking forward to a fully automated factory that produces not just one product, but hundreds, or even thousands, of digital serial manufacturing models.
Algorithm-based artificial intelligence is empowering the whole world, and the development of the 3D printing industry is facing a historic opportunity to take off!
If you know deeply, you can go far by doing. Based on a global network of manufacturing experts, 3D Science Valley provides the industry with an in-depth look at additive and intelligent manufacturing from a global perspective. For more analysis in the field of additive manufacturing, follow the white paper series published by 3D Science Valley.
Website Submissions l Send to [email protected]