A D V A N C E D M A T E R I A L S & P R O C E S S E S | J U L Y / A U G U S T 2 0 2 2 7 MACHINE LEARNING | AI PREDICTING BATTERY LIFE WITH MACHINE LEARNING Researchers at the DOE’S Argonne National Laboratory, Lemont, Ill., are using machine learning to predict the lifetimes of a wide range of battery chemistries. By using experimental data from a set of 300 batteries representing six different chemistries, the team can accurately determine how long various batteries will continue to cycle. The study relied on extensive experimental work done at Argonne on a range of battery cathode materials, especially the lab’s patented nickel-manganese- cobalt-based cathode. “We had batteries that represented different chemistries, that have different ways that they would degrade and fail,” says computational scientist Noah Paulson. “The value of this study is that it gave us signals that are characteristic of how different batteries perform.” Paulson believes the machine learning algorithm could accelerate development and testing of battery materials. “Say you have a newmaterial, and you cycle it a few times. You could use our algorithm to predict its longevity, and then make decisions as to whether you want to continue to cycle it experimentally or not,” he says. “One of the things we’re able to do is to train the algorithm on a known chemistry and have it make predictions on an unknown chemistry.” Further study in this area could potentially guide the future of lithium-ion batteries, adds Paulson. anl.gov. AI ASSISTANCE UP FOR DEBATE A team of researchers from Brookhaven National Laboratory, Upton, N.Y., the University of Liverpool in the U.K., and Ruhr University Bochum in Germany developed a new artificial intelligence (AI) agent called the x-ray crystallography companion agent (XCA) that assists scientists by classifying x-ray diffraction (XRD) patterns automatically during measurements. XCA uses a collection of individual AIs that are trained semi- independently of each other. Each agent has a slightly different weighting within its neural network. When presented with data, each AI “votes” based on its own interpretation and analysis. Once the AIs cast their final votes, the XCA approach uses a vote tally to interpret what the most likely atomic structure is and to suggest how confident the researchers should be of the AI analysis. Essentially, XCA is a group of AIs that debate each other while analyzing live-streaming x-ray data. Consensus among the ensemble implies confidence in the results because differing viewpoints still result in a common conclusion. However, strong disagreement can suggest that the analysis was poorly posed, and researchers should reexamine their assumptions. Unlike many other AI approaches in this field, this unique “ensemble voting” approach provides both predictions and uncertainties. In effect, this makes the approach a digital expert in XRD analysis. This approach demonstrates how AI and human researchers can work together to address scientific challenges such as developing new energy technologies and supporting human health. The study found that XCA can classify the materials as effectively as a human expert, but in fractions of a second. bnl.gov. A newmachine learning technique could reduce the cost of battery development. AI agents observe streaming x-ray data, argue among themselves, and vote to establish both classification and uncertainty in the prediction—offering an educated guess about the atomic structure of the material under analysis. Courtesy of BNL.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==