The Defense Advanced Research Projects Agency is awarding up to $10.3 million to a team led by the University of Michigan to investigate how long 3D-printed metal parts would last in the field.
Veera Sundararaghavan programs a 3D printer while PhD student Michael Philipchuk observes. Sundararaghavan is leading a DARPA project to predict the lifetimes of metal parts printed with laser powder bed fusion (LPBF). Printers that produce plastic parts, like the one in the photo, will be used to build mounts for the LPBF monitoring system. Image Credit: Marcin Szczepanski/Michigan Engineering
When military equipment malfunctions in remote regions, a part sent from the manufacturer can take weeks to arrive. 3D printing, notably laser powder bed fusion, or LPBF, is an expensive method of manufacturing parts, and they are unlikely to be as durable as cold-forged parts. However, downtime is more expensive in terms of lost work hours, thus military agencies would prefer the option of commissioning locally created parts or bringing their own 3D printers.
The difficulty is ensuring longevity. Military parts are subjected to rigorous testing, and the manufacturing process is typically so consistent that samples of one part produced by a specific machine reliably indicate how all such parts produced by that machine will perform. This is not the case with LPBF, where flaws in the material structure are more widespread and unpredictable.
LPBF involves hitting a metal powder bed with lasers until it freezes into the required cross-section. Then, additional powder is added, and the lasers fuse the next layer to the first from underneath. This continues until the portion is finished.
Depending on which model of LPBF printer you use, you might get different microstructures and different properties. The laser spot size and laser power levels might be different. The scanning strategies might be different. These things change the quality of the part.
Veera Sundararaghavan, Professor and Study Principal Investigator, Aerospace Engineering, University of Michigan
“Our aim is to guarantee the quality of the part as you print,” added Sundararaghavan.
Sundararaghavan and his partners propose meticulously recording the printing process and creating a digital twin of each part depending on the flaws that arise. The team will next simulate repeated loads on the component to see where fractures emerge and how long it takes.
These fatigue models can use actual service data to anticipate when a part will break. The team will validate these models using fatigue experiments. The four-year initiative is dubbed Predictive Real-Time Intelligence for Metal Endurance (PRIME).
To understand the lifespan of LPBF parts, we must push the current boundaries of the field and detect even the most critical defects that impact component performance. Through the PRIME project, we are doing exactly that—leveraging state-of-the-art monitoring and AI techniques to redefine what’s possible.
Mohsen Taheri Andani, Assistant Professor, Mechanical Engineering, Texas A&M University
Three partners—Addiguru, a monitoring firm for additive manufacturing, Texas A&M University, and the ASTM Additive Manufacturing Center of Excellence—will collaborate to create procedures and standards for data collection during LPBF manufacture. They will equip LPBF machines with an optical camera and two infrared cameras to capture near- and far-infrared signals that indicate where heat is accumulating in the sample.
Addiguru is pioneering multi-sensor integration using an acoustic sensor. Addiguru picked a sensor that was initially meant to detect birdsong, but it will now listen for the sounds of porosity problems in metal. These techniques will allow the researchers to spot flaws as tiny as 0.025 millimeters, and the sensor suite will be compatible with the majority of LPBF devices.
Multi-sensor data, combined with advanced analytics, will provide critical insights to part manufacturers. This project will enable a comprehensive, real-time assessment of part quality, helping manufacturers make informed go/no-go decisions with confidence.
Shuchi “SK” Khurana, Founder and Chief Executive Officer, Addiguru
Meanwhile, part of the U-M contingent will collaborate with the 3D-printing simulation business AlphaSTAR to utilize the data to create digital twins of the printed items. They propose to integrate AlphaSTAR’s superior physics-based modeling of the LPBF process with U-M’s microscale component structural simulations. The modeling and simulation of the microstructure will also assist the team in identifying residual stresses, or stresses built into the part that may later lead to its failure.
The microstructures of 3D-printed parts contain crystal grains that produce different properties across different directions, brittle structures known as intermetallic phases, and internal pores that are different from those seen in their conventionally processed counterparts. Microstructure modeling will offer important inputs for fatigue life predictions.
Lei Chen, Associate Professor, Mechanical Engineering, University of Michigan
Chen plays a key role in the microstructure modeling effort.
Finally, U-M researchers will collaborate with partners from the University of California, San Diego, to run uncertainty quantification models on top of the microstructure models, predicting the part's resilience over time by digitally testing how the metal responds to the stresses it will likely encounter on the job. To see if such forecasts are right, Auburn University will stress the metal parts to the point of breaking.
“If PRIME takes off, it’s like giving 3D printing a crystal ball—predicting the lifetime of LPBF parts across platforms and turning critical part production into a low-cost, distributed dream,” stated Sundararaghavan.
The project is financed by DARPA's Structures Uniquely Resolved to Guarantee Endurance (SURGE) program.