Optimization of Airless Tires Composed of Fiber/Rubber Composites for High-Speed Vehicles Using ANN and Computational Analyses

An artificial neural network is employed to predict the hyper-elastic mechanical properties of glass fiber/rubber composites used for airless tires. The training data are generated through FEA for representative volume elements. All data are split into training, validation, and testing sets in a ratio of 0.7:0.15:0.15. The comparison between FEA and ANN reveals an error margin of approximately 7.0%, indicating good accuracy. Additionally, the sensitivity analysis based on the response surface methodology is conducted to identify critical design variables influencing the shape and thickness of the spoke and tread of airless tires. An innovative airless tire model has been created using key design variables: fiber volume fractions in the spokes and tread, and thicknesses of the upper and lower spoke sections. The goal-attain multi-objective optimization is performed to optimize four stiffness of the tire. To validate the effectiveness of the optimization, a 3D FE tire model is constructed with optimal design parameters and subjected to deformation analyses to compute four types of static tire stiffness. The discrepancies in stiffness between the two methods range from 0.11 to 7.59%. Finally, the optimized model of the tire undergoes dynamic analyses to assess its vibrational performance, resulting in significantly decaying reaction forces.