Written by Tu Yanping
Editor / Huang Dalu
Design / Division
来源 / Automotive News，springwise
No one thought that AI would become the biggest winner of the 2024 Nobel Prize, and the two major prizes in physics and chemistry are related to artificial intelligence research.
For the first time, many people know that machine learning models are based on physical equations, while artificial intelligence is used to study the structure of proteins.
Before we know it, we have entered the era of AI
As electric vehicles grow in popularity around the world, concerns about battery safety are more pressing than ever. What happens when AI and batteries are combined?
Prevent thermal runaway
A key issue in the safety of power batteries is thermal runaway, which is triggered by unpredictable temperature spikes that can lead to catastrophic fires or even explosions in lithium-ion batteries.
New research from the University of Arizona offers a novel solution to this problem, combining machine learning with thermal sensors.
Principal investigator and PhD student Basab · Ranjan Das Goswami ·· and project principal investigator Professor Vitaliy Yurkiv · developed a system that can sense, predict and identify thermal runaway events within EV batteries.
Thermal sensors are wrapped around individual cells (up to 1,000 cells tightly packed together to form a complete cell), and these sensors are fed into machine learning algorithms trained on historical data.
The algorithm analyzes patterns to predict future overheating events and warns of potential failures. In an interview with Springwise, Goswami explained: "This approach enables real-time monitoring and early intervention, reducing the likelihood of catastrophic failure of an electric vehicle. ”
The approach used by Professors Goswami and Yurkiv is different from traditional approaches because of their innovative combination of AI and multiphysics models with lightweight sensors. This is more cost-effective than using bulky thermal imaging technology, and can be integrated into existing battery management systems to accurately predict temperature spikes in real-time.
As Goswami concludes, "This convergence of disciplines has allowed us to move from reactive security measures to proactive preventive measures." ”
The team received a $599808 grant from the Defense Incentive Competition Research Program of the United States Department of Defense. They are also exploring working with automakers to bring the technology to commercial use.
Find out the battery defects
Many tech companies are scrambling to find defects in power batteries that can lead to fires and other problems, and AI is helping them do just that.
They are training AI models to quickly assess what is normal and what is not. Automated tools greatly speed up quality checks.
Peter Kostka, Director of Battery Solutions at PDF Solutions, said at the Detroit Battery Show on October 10: "The key to AI is scalability: the more you deploy, the better. ”
PDF Solutions教它的AI模型来理解电池结构。
"If our model learns, 'Hey, these are the things I should normally see'...... Then the same model can be applied to different production lines, and we can get scalability. He said.
UnitX also uses AI to improve the defect identification process. Its 3D technology can identify subtle anomalies at high speeds. CEO Keven Wang said at the Battery Show that it can also detect deeper depths than 2D vision.
In the UnitX case study, a human operator scans a battery every 5 minutes, while the AI tool scans one every 3.5 seconds. According to the case study, a factory can reassign three human inspectors by using AI tools, Wang said.
"It takes looking at the flaws that have been seen before, but you'd be amazed at how good it is and how little sample is needed to teach," he said. ”
Accelerate development
Richard Ahlfeld, ·CEO of AI software provider Monolith, said battery companies have been using machine learning for years, but much of the battery industry has yet to embrace AI. He said the technology could cut battery testing time in half.
"The electric vehicle race has become more intense." "People are now thinking, 'Okay, what else can we do to speed things up?'" he said. And it's a tool that has been proven to speed up development considerably. ”
Monolith's AI software helps Jota Sport racing engineers optimize and validate track test and simulation data▼
NIO Europe said in September that it would use Monolith's technology to build a federated machine learning model that would be used to compare current vehicle field data with benchmark data. It will also reduce the time spent on battery data cleaning, resampling, analysis, and detection of anomalies.
Discover the next generation of materials
Many companies are using AI to predict and optimize battery health through the vehicle's battery management system; cleanup, classification, and reorganization through ChatGPT; and mapping molecules to discover the next generation of materials.
Alfeld said knowing health conditions can help drivers optimize charging and potentially extend battery life by 10 to 20 percent.
Hu Qichao, CEO of SES AI, said that SES AI is developing AI models to map more molecules than humans are able to draw. He said the models could become just as smart, if not smarter, than the top chemists.
SES AI believes these molecular maps will accelerate the discovery of materials that will solve any battery problem in electric vehicles, electronics, grid storage, and other applications.
But human scientists are the key to making databases work.
"Human scientists still need to synthesize models, use and actually test batteries. So it's almost like a model to create an idea, but the validation of the idea is still done by a human. He said.
The future of AI?
Patrick Hertzke, partner in the automotive and assembly practice at McKinsey &·Company's Future Mobility Research Center, said advances in chemical materials are where the most exciting potential of AI in batteries lies.
He said many companies are conducting incremental testing to improve batteries.
"It's like making a vaccine or making a drug. It's not easy, and it's not linear. Hertzke said. But based on breakthroughs in the pharmaceutical sector, "you should likewise be very excited about the potential for chemical improvements in the battery space." ”
Battery technology companies say that potential could be years away.
"Battery manufacturing is more of an art than a science." Manan Pathak, ·CEO of BattGenie, said at the Battery Show, "It's very difficult to have an end-to-end manufacturing process to make really good, repeatable batteries with very low error rates." ”
Even to catch defects, AI models need to be trained humans, Wang said.
"AI is another form of algorithm." "It's not a panacea. It's not magic. It predicts things very well. ”