Vehicle manufacturer BMW Group and French artificial intelligence company Mistral AI have launched a strategic partnership to advance the application of AI in crash simulations. Their cooperation is aimed at improving the quality, precision and speed of performing complex engineering tasks within car development. For the BMW Group, this represents an initial step towards the expansion of specific AI models to other areas of vehicle development, as well as to the entire value chain within the group.
Collision simulations
BMW points out that the use of industrial data is a key factor for turning artificial intelligence into new value. By combining large sets of technical data with Mistral AI’s model training capabilities, specialized artificial intelligence is created that directly supports complex development processes.
The scope and complexity of collision simulations in the BMW Group bring the need for artificial intelligence that is closely specialized in this industrial area. The company conducts thousands of virtual collision simulations on a weekly basis, generating huge amounts of data. Over the years, these processes have resulted in a historical set that exceeds a petabyte of data from collision simulations. This vast volume provides extremely detailed insights into vehicle structures and material behavior under stress, forming a unique basis for training an industrial AI model.
Large industrial models
In order to successfully scale this approach, BMW focuses on the so-called Large Industry Models (LIM). These are artificial intelligence systems that are trained exclusively on specific engineering and simulation data collected from vehicle development and safety testing. Unlike general-purpose AI systems, these models incorporate domain-specific knowledge directly into their structure.
The process of implementing such models requires not only the existence of industrial data, but also deep expertise and technical environments that enable AI systems to learn directly from BMW’s development processes. This collaboration, says Mistral, clearly demonstrates how models adapted to specific industries can help solve extremely complex engineering challenges, such as those in crash-test simulations.