Neural Network-Based Adaptive Clamping Force Correction Strategy for Electromechanical Brake Control

Electromechanical braking systems typically rely on pressure sensors to obtain feedback pressure values, but these sensors are prone to damage under high-temperature conditions. Therefore, the current study focuses on technologies that do not depend on clamping force sensors. These methods assume that the physical components of the braking system, such as the thickness and texture of the brake pads, remain unchanged. Under such assumptions, when the motor rotates to a specific angle, the brake caliper is expected to apply the predetermined clamping force. However, due to the gradual wear of the brake pads, their thickness will decrease over time. This reduction in thickness actually increases the real physical gap in the braking system. Under these conditions, even if the motor angle remains unchanged, the actual applied clamping force will decrease due to the increased physical gap, resulting in a clamping force that is less than that estimated by the motor angle, thus affecting the precision of clamping force control. To address this issue, a self-adaptive correction or compensation method is needed to predict accurate clamping forces under conditions of brake pad wear to ensure braking performance. Therefore, this study proposes a neural network-based adaptive clamping force correction strategy. This strategy utilizes a multilayer perceptron network, taking the clamping force estimated by the motor angle as input and the actual clamping force as the training label. The network outputs a predicted distribution of clamping forces using the softmax function. By calculating the expected value of this distribution, the accurate clamping force is predicted, thus achieving feedback control of the braking system and ensuring its stability and performance.