Optimization in airless tires design using backpropagation neural network (BPNN) and genetic algorithm (GA) approaches

Airless tires are designed and produced to overcome problems in the radial tires and solid tires. This tire provides a safe and comfortable driving experience in a vehicle during operation. Moreover, this tire still work when it hit by sharp objects (i.e., spike, nail, gun projectile, etc.). This research will be focused on designing airless tires using three parameters input, namely spoke thickness, rhombic angle, and rubber material. Each parameter uses three different levels, so the total design number is 27 designs. The thickness parameter of spoke levels was varied from 2 mm, 3 mm, and 4 mm, where the rhombic angles parameter was varied from 100°, 120°, and 135°. The last parameter (i.e., type of rubber material) was used in designing are Polyurathane L42, Polyurathane L100, and Polyurathane L135. The value of deflection and total stress every model are then calculated using finite element software. Furthermore, artificial intelligence using backpropagation of neural network (BPNN) was developed and utilized as a forecasting tool to predict the relationship between input (spoke thickness, rhombic angle, and rubber material) and output (deflection and total stress) of the airless tire models. Next, an optimization method using genetic algorithm (GA) is then employed to find the best design of the airless tire. Moreover, the best airless design will be selected to be produced as an airless-tire prototype.