Estimation of the braking torque for MotoGP class motorcycles with carbon braking systems through machine learning algorithms

This paper investigates new methods for estimating the braking torque generated by carbon brakes mounted on MotoGP class motorcycles. The operating characteristics of the brake is determined by taking into account that the friction coefficient between the pads and the disc depends on three main factors: the disc temperature, the disc-pads relative speed and the normal force applied. Firstly, a numerical model based on an optimal estimation algorithm is implemented to represent the thermal dynamics of the system. In particular, the monitoring of the brake temperature is performed through a Bayesian recursive filtering algorithm, in which the information measured by one sensor is combined with a prediction made on the basis of a physical model. Then, since the exact physical relationship between the three inputs and the friction coefficient is very hard to find by using a classical model-based approach, an artificial neural network is developed to approximate the unknown underlying mapping function from inputs to output.