The Toyota Research Institute invested $2.4 million in a research team at the University of Michigan, which will combine machine learning and data simulation to improve the prediction of battery properties.
Krishna Garikipati, professor of mechanical engineering and mathematics, is the lead investigator for the team, alongside Vikram Gavini, associate professor of mechanical and materials science engineering.
According to Gavini, the investment will take effect on May 1, with a chosen group of University students, scientists and scholars set to join the team.
One of the goals of the project is to develop a new type of battery that will be able to power vehicles that produce zero emissions, which aligns with one of the TRI’s main goals of reducing carbon dioxide emissions by 90 percent by 2050.
“Accelerating the pace of materials discovery will help lay the groundwork for the future of clean energy and bring us even closer to achieving Toyota's vision of reducing global average new-vehicle CO2 emissions by 90 percent by 2050,” Eric Krotkov, chief science officer at TRI, stated in a University Record article about the research.
To create a more effective battery, the team will work to combine artificial intelligence with the physics of materials, which will eventually enable them to predict material properties, Gavini said.
“We are solving equations at smaller scales and then we are collecting a lot of data on a small scale and then we are using artificial intelligence to figure out how is the materials physics at that particular scale and then passing it on to the next scale,” Gavini said. “The hope is that eventually, if we do this the right way, then we will be able to predict material properties.”
Duraisamy noted the significance of simulations in determining material properties, noting the use of the laws of physics and data can make the simulations more effective.
“The goal is to run more accurate simulations, and to run more accurate simulations you combine physical laws with data,” he said.
The simulations used in the research will include the ConFlux cluster, a computing platform that combines simulations with data sets to increase the speed of material development.
Duraisamy believes the simulations will complement the experimentation, allowing the research to be more effective.
“If you run the simulation, you can get any property you want, anywhere you want,” he said.
Duraisamy also said the simulations will make the experimentation more expansive.
“If you can use physical laws along with data, then we can make these simulations much more affordable and realistic,” he said.
Gavini said he appreciated the investment, specifically acknowledging the TRI’s emphasis on fundamental research that can expand across multiple disciplines.
“The Toyota Research Institute is funding fundamental science problems, which is very unlike what may happen to other industrial investments in the University,” Gavini said. “They are not hesitant to fund high-risk problems and I think that’s very good for the University, because at the University we work on such problems.”
Duraisamy echoed the importance of emphasizing fundamental research, stating it enables more widespread results.
“There are many different aspects to this, and since we are doing basic or fundamental research, the implications can be across the board,” he said.
The TRI also invested in research projects at Stanford University, the Massachusetts Institute of Technology, the University at Buffalo, the University of Connecticut and Ilika, a United Kingdom-based material science company.
Gavini recognizes the importance of the project, acknowledging the benefits to the University.
“This is a significant investment that Toyota is making in the University of Michigan and in our group, and this will continue going forward,” he said. “Hopefully it will be a successful project and we will have a very fruitful collaboration with these guys down the line.”