Much has been written about the sustainability aspects of additive manufacturing. It addresses the challenges traditional methods pose regarding material use, energy, and scrap production, amongst others. Today, it is also one of the areas of manufacturing in which artificial intelligence has gained a foothold. The synergy between the two is opening up untold opportunities.
Artificial intelligence, it would seem, is suddenly everywhere. And, if the experts are to be believed, this is just the start. Broadly speaking, artificial intelligence – AI - refers to computer systems that are capable of performing complex tasks that traditionally only humans could do, such as recognising speech and reasoning, making decisions, or identifying patterns and solving problems. As used today, AI comprises a broad range of technologies that include machine learning, deep learning, and natural language processing (NLP). In manufacturing, these technologies can support and optimise manufacturing processes, amongst others, through improved data analysis and decision-making. And while manufacturing may seem to lag somewhat behind other sectors in adopting the use of AI, this is due to change. The global artificial intelligence market for manufacturing, estimated to be worth $3.2 billion in 2023, is projected to reach a value of USD 20.8 billion by 2028, said the authors of a recent report published by Markets and Markets. This growth will mainly be driven by the rising need to handle increasingly large and complex datasets, emerging IoT and automation technology.
According to a recent MIT Technology Review Insights Report, compiled in collaboration with marketing intelligence company Dataiku, manufacturing companies can use AI to ‘streamline processes and fight downtime, adopt robotics that promote safety and speed, allow AI to detect anomalies quickly through computer vision, and develop AI systems to process vast volumes of data to identify patterns and predict customer needs’.
Specifically, therefore, there are five areas where AI can provide productivity gains in manufacturing, says the U.S. National Institute of Standards and Technology. These can be summed up as predictive maintenance; predictive quality; scrap reduction; increasing yield and throughput; and demand and inventory forecasting. One of the manufacturing technologies currently benefitting from AI technology is additive manufacturing. How?
There are different ways in which artificial intelligence can benefit additive manufacturing. In the past, one problem has been the high defect and scrap rates that have hampered the widespread adoption of AM processes in industry, combined with the fact that the complex shapes enabled by AM can be difficult and expensive to inspect after production.
Artificial intelligence can enhance quality assurance and defect detection through the use of vision systems to monitor the production process in real-time. Potential defects can be detected as they occur - even when not visible to the naked eye - reducing the number of potentially defective products produced. One effort in this area was, for example, a project called AMAI - Additive Manufacturing and Artificial Intelligence – conducted in 2021 by EIT Manufacturing, a public-private partnership co-funded by the European Union. AMAI aimed to develop and bring to market a novel solution for in-situ and in-line detection of geometrical errors and distortions in Powder Bed Fusion (PBF) while the part is being produced. EIT’s solution used the powder bed cameras available in most PBF systems but introduced a new combination of advanced image processing and data analytics to accurately reconstruct the geometry of each produced layer and detect possible geometrical deviations from the nominal shape. This made it possible to identify non-conformities in real-time, during production, thus ensuring consistent quality and reducing the amount of discarded material, scraps and the costs associated with traditional quality issues.
Technical maintenance company ATS also points out that in 3D printing, the use of AI can enhance quaality by ‘detecting potential defects and then helping operators or technicians to remedy the issue or — with the benefit of machine learning — allowing the manufacturing process to make those decisions on its own’. AI systems based on real-time control can, in this way, reduce the number of defective pieces that would otherwise be rejected and need to be disposed of.
Moreover, applied during the design and ideation stage, AI can help to streamline designs and to reduce process complexity. Not only can it be used to determine whether additive manufacturing offers the best choice to produce a particular part, but it also promotes a faster and more efficient design process, generating a design on the basis of a set of parameters or requirements. ‘AI in 3D printing design is also useful for topology optimization, using machine learning to render a design for the most efficient production possible’, writes ATS.
And by ensuring the optimum print parameters are determined prior to fabrication, far less time, effort and materials are wasted in testing and design iteration. If designers no longer have to make adjustments based on trial and error, major time, efficiency and cost advantages are created while at the same time optimising resource use and fostering a more straightforward quality assurance process.
One of the first companies to announce that it had successfully used generative AI to produce a lightweight 3D printed part – back in 2018 – was General Motors. In that year, the company announced a partnership with Autodesk, who had developed a program that combined cloud computing and AI-based algorithms to simultaneously generate multiple design solutions for a specific part. Design engineers could then select the one most suitable for production. Since then, AI is has been increasingly included in most of the 3D modelling systems used today.
GM used the technology to design a seat bracket with a fastening device for the seat belt that was 40% lighter and 20% stronger than the original part. In addition, the original part was composed of eight different components, now consolidated into a single 3D-printed part.
Likewise, AI has reduced the amount of trial and error in the material selection process. After all, the success of a 3D printed part depends strongly on selecting the right material that meets specific functional requirements. Use of AI-powered material databases can considerably facilitate this selection process. Designers can input their requirements in these databases, which then, leveraging machine learning algorithms, propose the material most fitting for the purpose. Machine learning models also allow predictions to be made regarding material behaviour and performance under different conditions. As a result, engineers can make well-informed decisions and select materials and design parts with higher confidence.
Advancements of this kind may also serve to reduce the need for extensive post-processing, thus considerably shorten the manufacturing time required. As well, automated post-processing solutions are being developed to streamline and improve the finishing process.
AI can also play a role in predictive maintenance of the printer itself by monitoring the lifecycle and by predicting maintenance needs based on historical data. In fact, predictive maintenance is an ‘enabler of achieving a sustainable production in 3D printing processes’, according to a paper published last year by the Fraunhofer Institute for Production Technology IPT. In the paper, the researchers explored how predictive maintenance could impact Fused Deposition Modeling of the high-performance polymer PEEK, prompted by the fact that the aging effects of the 3D printer’s nozzle have a direct negative impact on product quality but are not detectable from the outside. Using machine learning, they built a regression model through which it was possible to predict the remaining useful life of the nozzle, ‘enabling replacements to be made in a data-driven way, thus improving the maintenance strategy,’ wrote the scientists.
Scratching the surface
As a technology that is already accustomed to using computational modelling and simulation, AM is well-positioned to reap the benefits that AI has in store, especially with the emergence of unsupervised machine learning models. With unsupervised machine learning, AI learns from the data environment it is placed into, rather than by being force-fed labelled datasets that train algorithms to predict outcomes and recognise patterns. AI may well be the transformative technology AM needs to provide the quality, repeatability, speed and efficiency required to boost its economic competitiveness. Dogged by variability and uncertainty, AM technologies are still striving to achieve the level of consistent manufacturing of products - with guaranteed high quality – that is required in high-performance sectors such as aerospace or the medical industry. AI, with its data-driven decision-making capabilities and ability to detect anomalies and optimise processes, has the potential to take AM to the next level.
- https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-manufacturing-market-72679105.html?gad_source=1&gclid=Cj0KCQiA84CvBhCaARIsAMkAvkJZjFDfQqMul4CuqfvW-64lLjtzuePingR0Q0j-ENb6nyfbZh-4X2UaAvPyEALw_wcB
- https://3dx.ae/blogs/2023/08/ai-in-optimizing-3d-printing/ (https://www.advancedtech.com)
- Machine Learning Pipeline for Predictive Maintenance in Polymer 3D Printing, Henrik Heymanna, Robert H. Schmitta, https://doi.org/10.1016/j.procir.2023.03.058