Additive manufacturing processes continue to improve thanks to innovative efforts and growing adoption. This week marks a major milestone for the industry as a group of researchers from the University of Illinois Urbana-Champaign published a paper in the Journal of Intelligent Manufacturing that demonstrates a new method of training deep machine learning models to detect defects more effectively. Here”s what you need to know.
Additive Manufacturing is the Future
The 3D printing market will continue to expand, according to a study put out by BCG. The report predicts the additive manufacturing market will reach +95B by 2030. This growth is being fueled by lower costs, more availability, reduced technical barriers, and AI integration.
Notably, today’s 3D printing processes and devices can create more complex components than their predecessors. These devices enable engineers to develop never-before-seen innovations. Of course, as the prints get more complex, it becomes harder to detect defects, especially when discussing prints with internal mechanisms.
Reduced Failure in Additive Manufacturing Processes
It’s one thing when a 3D print fails on a consumer item. It may cost the company and the consumer funds and time. It’s another scenario when a 3D-printed piece is a vital component in aerospace or medical applications. Here, the stakes are much higher, and quality control has to be exceptional.
Vital to Quality
When printing complex three-dimensional shapes that have certain internal features, it can be impossible to do a proper quality check without creating massive delays and using X-ray computed tomography (CT) scans. In the past, this issue has resulted in 3D-printed components being limited in certain applications.
Source – University of Illinois Urbana-Champaign
These scans are costly and time-consuming. As such, there’s a huge demand for a more reliable and effective strategy to determine when a 3D print has defects. Researchers determined they could leverage the vast wealth of CT scans available via databases to help create a library of potential fail characteristics.
A New Study Reveals a Better Option
In their paper “Detecting and classifying hidden defects in additively manufactured parts using deep learning and X-ray computed tomography,” researchers explain how they programmed a highly effective AI algorithm that could accurately spot defects within prints without the need to use X-rays. Their study has shown impressive results, providing hope that in the future, 3D additive manufacturing could become the industry standard.
Printing the Test Parts
The team needed to create complex prints that could hold up under pressure and heat. As such, they determined that a nozzle with 3D internals would be the best test subject. Interestingly, 3D-printed nozzles are very popular because they can have specific design traits that help to navigate flow and alter pressure under certain scenarios.
Specifically, the team created 227 resin-based AM nozzles for the study. From the observable eye, all of these units looked pristine. However, some of the nozzles had purpose-built defects that would later be used to determine if the algorithm functioned properly. These defects were created to sit deep within the nozzle, ensuring that the study was difficult for the AI to pass.
Deep Machine Learning
The researchers leveraged an AI approach called deep machine learning. This style of learning uses computer simulations to replicate millions of scenarios. Each scenario produces unique data that helps the AI learn to recognize the situation in the future and how to react.
In this study, engineers began by feeding the AI algorithm the prominent online defeat CT scan data. They then scanned a series of prints to create 100,334 cross-section image slices. This collection of images made the first control group. Notably, the group ensured a 13.6% defect rate in the prints.
Vision Transformer (ViT) Model
The next step was to begin the simulations. The engineers had the algorithm run 50K synthetic defect simulations based on the data they input. The errors include a diverse selection of shapes, sizes, locations, and depths. This data was then superimposed onto defect-free parts to create a comparison base.
Capable Solution
The AI algorithm then scanned the 227 3D parts for defects. Notably, none of the defects in the parts were the same as in the image used to program the algorithm. As such, the AI needed to make judgment calls based on its findings and data alone. It accomplished this task efficiently, marking a major advancement in the sector.
Results
The results of this study have excited many in the industry. The Ai algorithm determined with 90% accuracy which parts had defects and which were printed correctly. Impressively, the AI even found defects that other AI models currently in use in the industrial sector had examined and missed.
Benefits that Deep Learning AI Brings to Additive Manufacturing Processes
There are a lot of reasons why this research can revolutionize the sector. Deep learning protocols have already helped to drive efficiency and innovation in other sectors. Now, this advanced tech can be used to speed up data modeling, improving safety while helping to create more advanced deep-learning models that can keep up with the complex printing processes in use today.
Accuracy
One of the biggest advantages of this research is the 90% accuracy rate. For manufacturers, having an AI determine 90% of the defects in your production line is a major upgrade. This algorithm will eliminate the need for CT scans on only the most critical components. As such, this approach will drive innovation as more complex designs can be created and tested via simulations.
Lower Cost
This deep learning AI model will lower the cost and time needed to train advanced AI models to detect 3D printing errors. These models will continually improve as more data becomes available. As such, a 90% success rate is a great place for the researchers to start. Future innovations could drive that rate up towards 99%, lowering costs for manufacturers across the board.
Fully Automated Systems
Deep learning models combined with additive manufacturing processes allow for the full automation of the creation process. Already, there are industrial metal 3D printers creating products for advanced operations. These devices have not been great at testing their creation in the past, which usually required a human.
The AI algorithm created by this team could be the solution to this problem as it would enable these systems to create and quality control their efforts internally. It also could be coupled with a recycling process. Imagine the AI recycling the defective pieces automatically when detected. Intelligent manufacturing systems like these will one day be at the forefront of the 4th industrial revolution.
Future of Additive Manufacturing
The future of additive manufacturing looks bright. These devices are found to be used across multiple industries, including aerospace, biotech, and medical. Here are some of the main reasons industrial 3D printing has become so popular over the last three years.
Lower Costs Units
One of the main reasons these devices are becoming more common is that they are now cheaper to own and operate. Manufacturers have more competition, which results in more innovation and lower prices. As such, businesses can integrate additive manufacturing into their models for less, improving efficiency and reducing waste.
Innovative Materials
Another major innovation is the creation of new printing materials. Composite and metals can now be used to create more durable components. Some additive manufacturing units can print complex designs that leverage multiple materials. These materials are lighter and stronger than ever. As such, they are sure to find more usability across industries.
AI Upgrades to Additive Manufacturing Processes
AI is one of the biggest reasons additive printing could become a daily task. AI models have been created for nearly every step of the process. Additionally, AI chatbots could make explaining your 3D print concepts as easy as sending a text. It’s already being used to help streamline the creation process and reduce waste.
This latest research integrates AI into the quality control back end. This integration was the final component needed to create a completely self-sufficient 3D printing ecosystem. In the future, these devices will create, print, and test their products without human intervention.
Bioprinters
One of the biggest innovations in 3D printing is the creation of bio 3D printers. These devices can print biomaterials such as cells and muscle tissue. These units are already in use in creating tiny soft robots that can navigate and organize into clusters to help complete difficult tasks such as water purification.
3D Printers in Space
Another cool innovation that has hit the additive manufacturing sector is printers capable of operating in microgravity environments. The device was created by the ESA (European Space Agency) and tested last week with success. Unlike its earthly counterparts, this unit needed to be completely enclosed to prevent heat or fumes from escaping.
In the future, space travelers will need a way to print what they need. They will be too far from Earth to rely on rocket ship deliveries, and there are sure to be scenarios when customized and never-before-used tools will become a necessity. Consequently, there is a lot of investment in the market from the aerospace sector.
Top Additive Manufacturing Firms
There are many additive manufacturers currently in operation. These firms are pioneering the future of 3D printing and helping to create a sustainable existance with less waste. Here are some businesses that could benefit from the defect research presented.
Stratasys has been in the 3D printing industry since 2010. Due to its quality solutions, the manufacturer has branched into the aerospace, medical, automotive, and consumer sectors with great success. Currently, the company has a popular software suite that improves design efficiency and deployment.
Stratasys stock has seen some ups and downs over its more than 14 years of trading. However, despite many market-rushing events such as COVID-19, Stratasys has prevailed. It remains a popular 3D manufacturing option and continues to expand operations. As such, it’s a wise addition to any portfolio.
Xometry is a new entry into the 3D printing market. The company, which provides a marketplace for 3D printing prototypes, manufacturing, and more, continues to expand its ecosystem. Today, Xometry boasts +5k suppliers and +8Kbuyers.
The firm’s unique blend of 3D printing on demand and the marketplace, which provides creatives and programmers with ROI opportunities, has many analysts predicting a bright future. For now, the stock has underperformed but has managed to position itself perfectly for the future onboarding of the traditional manufacturing sector, which could potentially be billions in revenue. As such, XMTR is worth checking out.
Researchers
The “Detecting and classifying hidden defects in additively manufactured parts using deep learning and X-ray computed tomography” study was conducted at the University of Illinois Urbana-Champaign. It was led by Professor of Mechanical Science and Engineering at Illinois Miles Bimrose, Sameh Tawfick, and William King. Additionally, Chenhui Shao from the University of Michigan, Davis McGregor from the University of Maryland, and Tianxiang Hu, Jiongxin Wang, and Zuozhu Liu from Zhejiang University participated.
Additive Manufacturing Processes Continue to Improve
Constant innovations are making the additive manufacturing market more competitive every day. In the future, 3D printers could find use cases everywhere. Thanks to the inventive minds behind the Vision Transformer (ViT) model, these products will be safer than ever.
Learn about other cool Additive Manufacturing Developments.