Challenge
New tire designs must satisfy both visual and performance-related requirements, such as driving dynamics and mechanical efficiency. Accelerating the design process while providing an environment where experts can finely control visual and functional elements is critical to improving overall product development efficiency.
Approach
An embedding-based learning model was developed to recognize specific tire components such as grooves and kerfs. A generative model, trained on internal tread pattern data and guided by design parameters, was used to create new pattern variations. To enhance model performance and training diversity, additional tire images were annotated and incorporated into the dataset for supervised learning.
Value Delivered
The system enabled rapid generation of diverse tread designs, significantly boosting the initial productivity of tire designers. By leveraging generative AI to explore previously unconsidered design possibilities—and integrating the solution with performance validation systems—the time required to develop a new tire was significantly reduced.
