Braking Force Identification of EMB Using Recursive Least-squares Method and Disturbance Observer Iteratively

An identification method using recursive least-squares method with moving data window and reduced-order disturbance observer iteratively is proposed in this paper to identify fast time-varying braking force in the electronic mechanical braking system (EMB). For the type of EMB which generates braking force by balls screw and motor mounted beside wheel, the actuator will go rapidly to eliminate clearance at beginning of braking process by means of raising the braking response speed, and at the same time, increasing the motor output torque which might be far larger than required. The proposed identification method is able to identify the point of contact between the brake pads and the disk in time by identifying the change of break force, and the torque of motor will be changed in time to reduce the braking force overshoot so that brake locking is avoided. Because of the existence of clearance in EMB system, when braking happens, braking force will change from zero to a big value at arbitrary moment, which enhances difficulty of online identification. The dynamics model of braking process is built in this paper, and the braking force is linear transformed to load torque of motor. RLS is used to identify the rotational inertia and resistance coefficient which is transformed to damping coefficient of the system. Then the identification parameters is used in reduced-order disturbance observer to identify the load torque, and then the RLS uses the load torque to identify parameters again. The above process is repeated iteratively. The data window is used to avoid the ‘data saturation’ and avoid the accumulation of error in the observer which can improve accuracy of identification result of time-varying parameters.