Brake Heat Capacity Prediction Using a Machine-Learned Friction Coefficient Model and Virtual Wheel Brake

Recently, the increasing complexity of systems and diverse customer demands have necessitated the development of highly efficient vehicles. The ability to accurately predict vehicle performance through simulation allows for the determination of design specifications before the construction of test vehicles, leading to reduced development schedules and costs. Therefore, detailed brake thermal performance predictions are required both for the front and rear brakes. Moreover, scenarios requiring validation, such as alpine conditions that apply braking severity to xEV with the regenerative braking system, have become increasingly diverse. To address this challenge, this study proposes a co-simulation method that incorporates a machine-learned brake pad friction coefficient prediction model to enhance the accuracy of brake thermal capacity predictions within the vehicle simulation environment. This innovative method allows for the simultaneous prediction of both front and rear-wheel brakes. The required brake torques for the front and rear wheels are calculated based on the vehicle model and driving scenarios. The brake system model generates the necessary pressure during deceleration, whereas the friction coefficient is crucial in creating brake torque, resulting in brake power for both the front and rear brakes. Within the simulation model, the virtual wheel brake calculates the speed, pressure, and disc temperature based on vehicle driving schedules. The machine-learned model utilizes these variables as inputs and returns the friction coefficient. The prediction accuracy of torque and disc temperature improved significantly as the virtual wheel brake utilized the friction coefficient received from the model trained using the mixed-effects random forest algorithm.