How Webasto will use AI to build better batteries

German roof and technology supplier Webasto has selected Monolith, a London-based artificial intelligence (AI) software provider, as its AI partner to accelerate battery validation and improve its electric vehicle battery testing.

Under the new partnership, Webasto will use Monolith’s AI software to enhance the efficiency of its battery module and pack testing.

Webasto develops batteries for a range of vehicle applications, building more than 500 battery system prototypes a year, thereby generating terabytes of raw data from thousands of tests. However, the generated test data is not in a suitable format for training AI models.

Monolith, meanwhile, builds AI software to process test data.

The Monolith software platform examines the test data to improve battery efficiency, quality and performance while reducing the time spent on time-consuming, repetitive physical tests. This frees up engineers to focus on other processes and shortens product development time.

Moreover, Monolith’s machine learning for test data can identify glitches and highlight the areas where processes can be optimized, thereby allowing Webasto to gain maximum insight from the data.

“We try to find anomalies,” Monolith CEO and founder Richard Ahlfeld told Automotive News Europe.

The partnership shows how AI can be harnessed to accelerate battery testing and validation processes. In other words, the partnership exemplifies how AI can reduce the cost of battery development and “get better products to market faster” as competition with China intensifies, Ahlfeld said.

“The speed of development in China is phenomenal. Meanwhile, European and American manufacturers don’t want to compromise on quality for speed. Our processes can help to reduce development costs and increase development speed without compromising on quality,” added Ahlfeld, who was a 2023 Automotive News Europe Rising Star.

Small job, big impact

Over the next 12 months, the partners plan to work on a variety of use cases. However, their initial focus will be on the development of bolted joints and the welding of battery modules, namely battery pack screw tightening.

“This process has a massive impact on the battery’s life,” Ahlfeld said.

Hundreds of parts are tested during the development and assessment of screw tightening torques and weld seams, generating huge amounts of data. A machine-learning monitoring system can capture the majority of errors during the testing process.

Monolith already has partnerships with other automotive companies to “develop higher quality products in half the time,” the company says.

For example, BMW Group uses Monolith’s machine learning software for vehicle safety in crash tests. Specifically, the software can more accurately predict how design changes could affect the rates of leg injuries in crashes and reduce the likelihood of those injuries happening. The machine learning software cuts the development time for BMW’s test engineers.