For the better part of a century, General Motors (GM) was the biggest automaker in the world. Now, amid a paradigm shift toward smarter, electrified vehicles, the leading American automaker is working to meet the moment – and to do that, it’s leveraging deep partnerships with U.S. National Laboratories, from their particle accelerators to their supercomputers. At a meeting of the Department of Energy’s Advanced Scientific Computing Advisory Committee (ASCAC) last week, Paul Krajewski – director of vehicle systems research for GM’s research and development center – highlighted the depth and value of GM’s work with the national labs to leverage HPC and advance vehicle technology.
Many of the national laboratory partners covered by HPCwire find the need for national laboratories only when they reach the limits of their own computational resources. But GM, Krajewski explained, fosters deep relationships with the national labs that pervade numerous facets of its research and development processes – not just by way of their high-performance computing equipment, but also experimental equipment, research expertise and software.
Krajewski pointed out that the national labs’ collaborations with GM have spanned many forms: standalone projects, contracts, partnerships, mutual participation in broader DOE contracts, facility use, discussions, paper publications. Throughout these collaborations, he said, GM has worked to leverage the national labs’ “excellent researchers” and “unique capabilities” – with an emphasis on data collection.
“One … opportunity I think is really important for this collaboration is the combined access to data,” he said. “The computing doesn’t give us the right answers if we don’t have the data – if we don’t have the ground truth and we don’t have the experimental results.”
Krajewski gave a high-level overview of where GM is investing its research money: “Battery technology, autonomy, fuel cell, advanced manufacturing,” with “almost half” of the funding devoted to battery technology and electrification.
“We are committed to an electric future,” Krajewski said. GM, he explained, had partnered with the national labs through a consortium called Battery500. That consortium – which loops in four national labs (led by Pacific Northwest National Lab) and five universities – aims to develop high-energy, rechargeable lithium-metal batteries for electric vehicles. (Lithium-metal batteries are distinct from lithium-ion batteries, offering substantially higher energy density but suffering from stability challenges.)
Through Battery500, GM has worked with the national labs to develop new processes and materials for lithium-metal batteries, heavily informing subsequent simulations and model development. “This is really important in generating the experimental data that we need to then create the models,” Krajewski said. “There’s also work now on multiscale modeling of lithium-ion batteries and then, ultimately, it will go beyond that to lithium-metal batteries, and this is where the computational capability is extremely important. If you want to be able to do these multiphysics, multiscale models that are very, very computationally intensive, you need to have that power that resides in the national labs.”
But GM isn’t putting all of its eggs in the battery basket – much of GM’s other research focuses on fuel cell technology, which would use hydrogen fuel cells to power electric motors, with only water and heat as byproducts. Krajewski said there was “a tremendous amount of ongoing work” with the National Renewable Energy Laboratory (NREL) on fuel cells, looking at both experimental capability and computational modeling. For instance, GM has been helping to fund the “H2FiLLS” hydrogen fuel cell filling simulation project, which has been making good use of NREL’s Intel-powered Eagle supercomputer (8 peak petaflops) to run computational fluid dynamics models of hydrogen storage tanks, mixing the simulation results with experimental testing to refine the simulation even further.
Similarly, the labs and GM have been working together on thermal management of electric vehicles. “Whether it’s thermal management of the batteries themselves or electric motors, it becomes a very important challenge,” Krajewski said, “and so there’s some ongoing work now on thermal management analysis with Argonne [National Laboratory] on motors and how [we can] optimize pushing these motors to their highest performance … without creating problems with overheating and how to manage the thermal stresses that we end up building up in these motors as we push them to the limits.” That work is powered by a 5 million core-hour allocation on Argonne’s Intel-powered Theta system (6.9 Linpack petaflops).
Manufacturing and lightweighting are also major focuses for GM. “When you look at the data that we have around manufacturing capability going forward, and when you think about a transition to chip production, mining, critical materials – all of these things are going to require manufacturing models so that we can be very efficient with it,” Krajewski said.
On the lightweighting front, GM has been working with the Advanced Photon Source (APS) particle accelerator at Argonne National Laboratory to work on characterization of so-called “Gen 3” steel, which aims to be stronger (and thus require less material for sufficient performance and safety in vehicles). Those experimental results from the APS have then been combined with finite element optimization on NREL’s (now decommissioned) Peregrine cluster. GM similarly partnered with Oak Ridge National Laboratory (ORNL) to develop a new aluminum alloy (DuAlumin 3D) via a combination of computational and experimental analysis.
“How do I use these materials once I’ve developed that?” Krajewski continued. “I now have a suite of materials I could apply; I have the performance requirements – in this case, … a body side structure with different parts on it; I could run a finite-element optimization using the cluster at NREL to optimize what material, what gauge, where I want to use it, and really come up with the best solution for applying these materials that I’ve also designed leveraging these techniques.”
And then, even later in the design process, GM is “targeting 100% virtual validation by 2025” to eliminate unnecessary dies, tools and prototype parts and to speed up vehicle development. “When we have those material models, now [you] can do things like predicting performance, predicting formability – can I make the shapes that I wanna make? – static and dynamic principles, crash performance,” he continued.
“These collaborations are essential to drive technology development,” Krajewski concluded. “There’s a lot of models for that collaboration, and combining the capability of experiments with this computational capability is the key to these collaborations and really moving technology forward.”