AI Surrogate Model for Tire Performance Prediction – DUNLOP (Company name: Sumitomo Rubber Industries, Ltd. [1] ) and Fujitsu Limited [2] today announced that they have jointly developed a technology, an AI surrogate model, to predict tire performance with high accuracy and in a short time using AI, confirming its effectiveness in a recent proof of concept. This technology, developed as part of DUNLOP’s long-term management strategy for design digital transformation, was applied to the structural analysis of tire deformation when in contact with the road surface. As a result, the analysis time was significantly reduced by approximately 90% from about 45 minutes to about 5 minutes, while achieving analysis of approximately 600,000 elements (meshes).
Based on the results of this proof of concept, both companies will proceed with the development of a design support tool for tire design, aiming for its practical implementation at DUNLOP by April 2027. This will enable DUNLOP to accelerate data-driven development and swiftly supply high-quality tires with enhanced safety and environmental performance to the market.
This technology is designed to run on “FUJITSU-MONAKA,” [3] a next-generation Arm-based CPU developed by Fujitsu that pursues both high performance and energy efficiency. Moving forward, both companies aim to optimize inference speed, accuracy, and power efficiency by commencing verification of this technology on a prototype of FUJITSU-MONAKA by December 2026.
Background
In manufacturing, CAE (Computer Aided Engineering) analysis [4], which simulates the behavior of products and structures to evaluate performance and safety, requires a substantial amount of computational time due to the increasing performance and complexity of products.
In tire design, FEM (Finite Element Method) analysis [5], a type of CAE analysis, is commonly used. While increasing the number of elements by refining the mesh improves accuracy, it also increases computation time and associated development costs, necessitating a balance between accuracy and computational load. Furthermore, analysis requires specialized knowledge, and securing skilled engineers is also a challenge.
To address these issues, both companies developed an AI surrogate model that rapidly predicts solutions to the governing equations of FEM using accumulated FEM analysis results as training data.
Proof of Concept Results
Leveraging DUNLOP’s tire design expertise and actual design data, along with Fujitsu’s AI technology, both companies jointly developed an AI surrogate model based on the Graph Neural Network (GNN) [6] algorithm. They conducted a proof of concept for tire structural analysis, focusing on evaluating deformation behavior and contact characteristics such as contact shape and pressure distribution when a tire is in contact with the road surface.
As a result, computation time for the analysis was reduced from approximately 45 minutes to about 5 minutes. The technology predicted the contact shape between the tire and the road surface with a high average accuracy of 87.7% compared to FEM analysis. This technology will enable faster decision-making and optimize costs, in addition to improving performance, by allowing the determination of tire structure and material specifications in fewer processes and a shorter time, which previously required multiple design processes.
A part of these results will be presented at the 31st Computational Engineering Conference, held from June 3, 2026.
Figure1: Proof of Concept Image
Figure 2: Relationship between accuracy and computation time in FEM analysis
Future Plans
Both companies will begin verification of this AI surrogate model using a prototype of FUJITSU-MONAKA by December 2026, aiming to optimize inference speed and power efficiency. They will also expand the application range of tire structural analysis and develop a design and development support tool that can be directly used by designers without specialized knowledge. DUNLOP aims to commence practical operation by April 2027.
DUNLOP aims to realize its vision of “Continuing to Provide ‘New Experiential Value’ Born from Rubber to Everyone” under the Long-term Corporate Strategy “R.I.S.E. 2035.” [7] Through this co-creation with Fujitsu, DUNLOP will further enhance its proprietary Rubber and Analytical Technology, strengthening its technological capabilities and accelerating innovation. By doing so, DUNLOP will put into practice its Purpose: “Through innovation we will create a future of joy and well-being for all.” and continue to deliver new value to society.
Fujitsu will leverage this initiative to promote horizontal deployment to large-scale FEM analysis in the automotive and other manufacturing industries. Moving forward, Fujitsu will contribute to optimizing development and promoting carbon neutrality through improved power efficiency in the manufacturing industry by developing an AI inference platform combining FUJITSU-MONAKA and GNN, and offering it on the AI platform “Fujitsu Kozuchi.” [8]
Trademark Information
All company or product names mentioned herein are trademarks or registered trademarks of their respective owners.
For more information visit: global.fujitsu
