Abstract: For a direct-drive linear AC servo system, a sliding mode variable structure control scheme based on fuzzy self-learning is proposed. Fuzzy self-learning is introduced into conventional sliding mode control, effectively reducing jitter caused by sliding mode switching control without affecting the overall robustness of the sliding mode control system. Simulation results show that the scheme has strong robustness to system parameter changes and load disturbances, and jitter is significantly suppressed. The system exhibits good dynamic and static performance. Keywords: Sliding mode variable structure control, fuzzy control, learning algorithm, AC linear servo system, direct drive Introduction Permanent magnet linear synchronous motors (PMLSMs) eliminate the indirect mechanical conversion device from rotary motion to linear motion, making them the best choice for actuators in high-precision, micro-feed servo systems. However, since the linear motor's mover directly drives the load, load changes and external disturbances directly affect the performance of the servo system. Parameter changes of the motor under various operating conditions and the nonlinearity of the motor itself will adversely affect the performance of the servo system. Traditional PID control cannot meet the requirements of high-performance linear direct drive AC servo systems. This paper proposes a new sliding mode variable structure control strategy based on the characteristics of the controlled object and the defects of traditional control strategies. Sliding mode variable structure control is a special discontinuous nonlinear control method. It is invariant to parameter changes, external disturbances and system uncertainties, and has the advantages of speed and simple implementation. However, the chattering phenomenon generated by the system affects the stability of the system operation. The common method to eliminate chattering is to replace the switching control with continuous saturated nonlinear control to smooth the discontinuous variable structure control. This method can eliminate chattering, but it also eliminates the anti-perturbation property of the sliding mode variable structure control system. Reference [5] proposes a method to control the sliding mode approaching law. Without affecting the anti-perturbation property of the system, chattering is weakened by controlling the speed at which the state variable approaches the sliding mode switching line. However, the specific approaching speed is difficult to determine in the actual system, and the dynamic and static characteristics of the system are difficult to achieve simultaneously. Reference [6] introduced fuzzy theory to eliminate chattering, but establishing accurate fuzzy rules is still quite difficult. This paper proposes a fuzzy self-learning sliding mode control strategy to learn online the uncertainties of the actual system (parameter changes and external disturbances, etc.). Through fuzzy inference, the direction and amplitude of the sliding mode switching control {6I} are adjusted in real time. Without affecting the robustness and fast tracking performance of the system, the impact velocity of the system motion state when passing through the sliding mode line is reduced, thereby fundamentally and greatly weakening the chattering of the system. [Click for details]