Abstract: This paper proposes an adaptive feedback linearization control scheme for AC servo systems. By applying online estimated parameters to coordinate transformation and nonlinear state inversion, linearization control is achieved. The problem of enhancing parameter convergence is discussed. Theoretical derivation and computer simulation both prove that this scheme has strong robustness to parameters and practical application value. Keywords: Inversion linearization control; Permanent magnet synchronous motor; Linearization convergence 1 Introduction An AC servo system consists of an AC permanent magnet synchronous motor (PMSM), position and speed sensors, current sensors, transistor inverter PWM, and control circuits. The servo motor is a nonlinear element because it contains a product term of electrical angular velocity and current i or i₀. Due to the influence of load disturbances and other factors, the study of robust control strategies for servo motors is a necessary task. 2 Model Linearization of AC Servo Motors For servo variable speed drive systems composed of PMSMs, it is convenient to describe and analyze their steady-state and dynamic performance using reference coordinates fixed to the rotor. Taking the direction of the permanent magnet's base magnetic field as the d-axis, and the electrical angle leading the d-axis by 90° along the direction of rotation as the q-axis, the spatial coordinates of the rotor reference coordinates are determined by the electrical angle θ between the q-axis and the fixed axis (A-phase winding axis). The rotational speed of the rotor reference coordinates is the shaft speed. Under the above assumptions, the voltage equation and torque equation expressed in rotor reference coordinates are as follows: where and represent the voltage and current of the d and q axes, respectively; is the component of the flux linkage in the d and q axes; is the rotor speed; Rs is the stator phase resistance; Td is the load torque; J is the total moment of inertia of the rotor and its load; B is the coefficient of viscous friction; and pn is the number of pole pairs of the motor. Combining equations (1), (2), and (3), we can obtain the mathematical model of the servo motor. 3. Adaptive Feedback Linearization Control Due to the uncertainty of load disturbance and stator resistance, the parameters N and V in the nonlinear state feedback cannot be equal to the actual parameters, thus destroying the condition for complete linearization and causing the complete linearization control strategy to fail. Therefore, we should study the problem of making the system completely linear and robust for the parameters N and V. Let the system (6) be expressed as and have the option of nonlinear coordinate transformation, where the adaptive estimation law of the parameters is obtained. Applying the estimated parameter values ​​to equation (10), we obtain the adaptive feedback linearization control law of the servo motor. 4. Convergence of Feedback Linearization Control 5. Computer Simulation Figures 1 and 2 show the directional control process of the flux linkage. As can be seen from Figure 3, although the initial estimated value of the parameters is not equal to the true value, with the help of the adaptive estimation of the parameters, the estimated value of the parameters can gradually approach the true value, thereby achieving the purpose of asymptotic linearization control. Figures 4 and 5 show that the d-axis flux linkage asymptotically approaches the reference flux linkage value, while the φ-axis flux linkage asymptotically approaches zero, thus achieving flux linkage orientation control. Figure 6 shows the speed response process, demonstrating that the operating speed can quickly and accurately reach the given speed. 6. Conclusion Through adaptive parameter estimation, the estimated parameters asymptotically approach the true values. This achieves fully linearized control based on nonlinear state feedback, realizing three-loop control of speed, rotor flux linkage, and coordinate axis orientation. Theoretical derivation and computer simulation both demonstrate the feasibility and effectiveness of this method in AC servo systems.