Machine learning aids development of automotive body structures

Designing automotive body structures is no simple task. To optimize these structures, designers of automotive parts must consider several factors, including crashworthiness, light-weighting, manufacturability, and local requirements.

While engineers often use their former experiences to associate certain structural configurations, feature shapes and patterns with particular performance abilities, it is difficult to analyze human knowledge in the form of general guidelines. Thankfully, machine learning and AI may provide a solution.

A recent project headed by Prof. Jami Shah and SIMCenter researcher Satchit Ramnath alongside researchers and students from Digital Design and Manufacturing Lab investigated the use of machine learning to extract automotive structure feature attributes and patterns associated with specific performance parameters.

“The methods developed allow us to generate data sets of sizes and structure that enable research on machine learning and artificial intelligence methods for CAD data, with the goal of supporting engineers in the virtual design process,” Ramnath said.

The knowledge extracted is used to develop superior structure designs that meet or exceed requirements. The ability to extract feature attributes and develop new models using machine learning provides researchers with a starting point in the design process that is already superior to old designs.