Scientists seek AI-driven metal 3D printing
Metal Tech News - March 22, 2023
Last updated 4/16/2023 at 7:12am
U of T researchers begin to compile database for self-printing additive manufacturing.
Tapping into the unlimited potential of additive manufacturing, University of Toronto researchers are reevaluating the processes of metal 3D printing to build the framework of a future system where production with printers is completely automated, down to each molecule.
Experimenting at the university's first metal 3D printing laboratory, a team led by Professor Yu Zou in the Faculty of Applied Science and Engineering at U of T is exploring the fundamental physics behind the 3D printing process, a study that hopes to set a foundation for future development and possible standardization in the limitless additive manufacturing process.
"We are working to uncover the fundamental physics behind the additive manufacturing process, as well as improving its robustness and creating novel structural and functional materials through its applications," said Zou.
Different from conventional manufacturing, which constructs components or parts from bulk materials, the metal 3D printing process enables microstructure and materials constitutions to be locally tailored, meaning they exhibit unique and distinct properties.
"For example, medical implants require human bone-like materials that are dense and hard on the outside, but porous on the inside," said Xiao Shang, a doctoral candidate in Zou's lab. "With traditional manufacturing, that's really hard to accomplish – but metal printing gives you a lot more control and customized products."
Typical manufacturing processes utilize molds to form a general shape and subtractive manufacturing to shave down the part into its precise arrangement, such as with a CNC (computer numerical control) machine. Additive manufacturing, by contrast, builds new objects by adding layers of material.
Not only does this process significantly reduce production time, but it also saves considerably on resource material and therefore cost, as well as energy consumption.
This is especially important when producing objects for the aerospace industry, like engine components, tooling parts for automotive production, critical mechanisms for nuclear reactors, and biology-friendly joint implants, all of which factor costs to an acute degree.
Reassessing 3D printing tech
The university's metal 3D printers are designed to specialize in two conventional additive manufacturing techniques – selective laser melting (SLM), a process that utilizes high-power lasers to fully melt and fuse metallic powders from the bottom up, and directed energy deposition (DED). Instead of seeing a shape emerge from a bed of powder, DED involves powder or wire extrusion that feeds into an adjacent superheated point, forming the shape upon a surface.
Typically, computer-aided design or CAD software is used to create a 3D model of the object and its layers. This technology has progressed so much that some firms have implemented methods in which a real-world object can be scanned to hasten the process instead of painstakingly creating a design from scratch.
Once a model is designed, a 3D printer gets to work and constructs the object layer by layer. No matter the technique, metal 3D printers function in much the same way as their polymer counterparts.
After the molten metal layer solidifies, it adheres to either the previous layer or the base plate called a substrate. Once each layer is complete, the machine will induce powder doping (a technique that incorporates particles that can change the composition of the base material) and laser melting until all layers are printed, and the object is completed.
"Conventional manufacturing techniques are still well-suited for large-scale industrial manufacturing," said Tianyi Lyu, a doctoral candidate under Zou. "But additive manufacturing has capabilities that go beyond what conventional techniques can do. These include the fabrication of complex geometries, rapid prototyping and customization of designs, and precise control of the material properties."
For example, dental professionals can use SLM to create dentures or implants customized to specific patients via a precise 3D model with dimensional accuracy that is within a few micrometers.
Rapid prototyping also allows for easy adjustments of the denture design, and since implants can require different material properties at distinct locations, this can be achieved by simply changing the process parameters.
The team is also applying novel experimental and analytical methods to gain a better understanding of the SLM and DED printing processes.
Foray into material composition
Currently, the researchers are focused on advanced steels, nickel-based superalloys, and high-entropy alloys. Furthermore, it is believed that they may expand to explore titanium and aluminum alloys in the future.
"One of the major bottlenecks in conventional alloy design today is the large processing times required to create and test new materials," said Ajay Talbot, a master's student in the materials science and engineering department. "This type of high-throughput design just isn't possible for conventional fabrication methods."
With additive manufacturing techniques such as directed energy deposition, the group is rapidly increasing the number of alloy systems explored, as they are able to alter the composition of materials during the printing process by adding or removing certain elements.
"We are also working towards intelligent manufacturing. During the metal 3D printing process, the interaction between a high-energy laser and the material only lasts for a few microseconds," said Jiahui Zhang, another doctoral candidate in the science and engineering department. "However, within this limited timeframe, multi-scale, multi-physics phenomena take place. Our main challenge is attaining data to capture these phenomena."
Zhang goes on to say, "In our research, we have successfully customized specific machine learning methods for different parts of the metal additive manufacturing lifecycle."
In the lab, high-speed infrared camera systems are integrated directly into the metal 3D printers. The team has also constructed an in-situ monitoring system based on the images taken by the printer to analyze and extract the key features of printed objects.
"With the development of computer vision, a well-trained deep learning model could automatically accomplish some basic tasks that human visual systems can do, such as classification, detection and segmentation," added Zhang.
University of Toronto Engineering
With the persisting issue of uncertainty regarding part integrity, the team has opted to actively work to apply machine learning and computer vision to develop a fully autonomous closed loop-controlled 3D printing system that can detect and correct defects that would otherwise emerge in parts made via additive manufacturing.
"Implementing these systems could greatly widen the adoption of metal additive manufacturing systems in the industry," said Zou.
Since building up the lab's metal printing capabilities, Zou and team have established partnerships with government research laboratories, including National Research Council Canada, and many Canadian companies.
"Metal 3D printing has the potential to revolutionize manufacturing as we know it," Zou added. "With robust autonomous systems, the cost of operating these systems can be dramatically reduced, allowing metal additive manufacturing to be adopted more widely across industries worldwide. The process also reduces materials and energy waste, leading towards a more sustainable manufacturing industry."