To enhance the grinding quality of spiral bevel gears, an intelligent control model for the grinding process of automotive helical conical gears based on force feedback has been designed. This model outputs the control voltage for the machine tool's permanent magnet synchronous motor (PMSM), ensuring that the motor speed constantly tracks the desired value. By adjusting the grinding generating speed, the grinding force is controlled, and the tooth surface roughness is reduced. Firstly, the state equation of a permanent magnet synchronous AC servo motor is established. By employing the second method of Lyapunov, an RM adaptive control algorithm is developed. It is found that the model output can efficiently track the reference model (RM) and adjust to variations in torque due to load. To further enhance the controller, a generalized regression neural network (GRNN) was developed; subsequently, training data were generated using the output voltage of the RM self-adjusting controller to achieve velocity regulation of the machine tool's servo motor. Finally, the results indicate that the GRNN controller is superior. It uses RM self-adjusting control data as samples for regression analysis, outputs control signals, and controls the angular velocities of each axis of the machine tool to control the grinding force within a reasonable threshold range, reducing the complexity of the controller and achieving lightweight. At the same time, the feasibility of the controller has been experimentally verified. This improves the roughness of the tooth surface during the grinding of spiral bevel gears and enhances the quality of vehicle operation.