Sumitomo Rubber and Fujitsu develop AI model to cut tire structure analysis time by approximately 90%

Sumitomo Rubber and Fujitsu jointly developed an AI technology called 'AI Surrogate Model' that predicts tire performance with high precision in short time. In a proof-of-concept using the model to predict tire deformation behavior when in contact with road surface, analysis time was reduced by approximately 90% compared to conventional methods—completing in about 5 minutes what previously took about 45 minutes at a scale of 600,000 elements (mesh).

The model uses Graph Neural Network (GNN) algorithm to infer finite element method (FEM) approximations based on input parameters such as tire mesh shape, air pressure, load, and material information. The prediction accuracy for tire-road contact shape was 87.7% on average compared to FEM. The model was designed to run on FUJITSU-MONAKA, Fujitsu's next-generation Arm-based CPU.

Sumitomo Rubber and Fujitsu aim to launch a design support tool based on this technology by April 2027. The companies plan to begin verification on FUJITSU-MONAKA test machines by December 2026. Fujitsu also plans to expand the technology's application to large-scale FEM used across manufacturing industries including automotive.