The Research On Auto-tuning Technology Of AC Servo System

The performance of AC servo system is closely related with its control parameters. Therefor adjustment of the parameters is necessary for satisfying control performace. Morever, allowing for requirements in different applications and condition variations, adaptive capability is necessary for servo systems, adjusting control parameters online according to variations in load's characteristic, and keeping the performance optimized. Researches on self-tuning strategies in AC servo system are urgent requirements for improving servo system's performance and efficiency.With the implementation of servo system parameter self-tuning as the thesis'primary target, mathematic models of the control loops are constructed from inside to outside according to the cascade inner structure of the servo system. Controller of each loop is designed and parameterized with separated self-tuing rules so that theoretical foundation is derived.A digital AC servo system platform is constructed for further simulation and experiment. A speed predictor based on linear recursive smoothing method is proposed, improving the speed detection effectively. Experiments are carried out for the verification of parameter adjustments in current, speed and position loop.Identification of the target permant magnet synchronous motor (PMSM)'s electric parameters is necessary for self-tuning of the current loop, therefor this thesis proposes an online parameter identification scheme based on adaptive observer. A correction of current feedback observing error is introduced in addition to the traditional dq current observer, with the feedback gain designed based on static theory. Adaptive law for parameter identification based on Popov's hyperstability theory is also derived. Further study is performed towards dead-time effect which could deteriorate identification precision and a novel compensation method is also proposed. The studies above achieve PMSM's electric parameters identification, laying foundations for the parameters self-tuning of the current loop.Identification of the mechanical parameters is necessary for determination of parameters in speed loop and position loop. Modifications have been made on traditional acceleration-deceleration inertia identification method based on track planning, improving its practicability. Detailed study is carried out for least square method (RLS) and forgetting factor least square method (FFLS) when applied to moment of inertia identification. A reduced order disturb torque observer is also proposed and its gain is determined with guaranteed stability. These researches achieve completed identification of mechanical parameters in the servo system, laying foundations for the parameters self-tuning of the speed and position loop. Self-tuning of the parameters of servo system is achieved based on identification of electric and mechanic parameters. In case of very large load inertia the largest possible bandwidth of the speed loop is derived, optimizing the parameter self-tuning of speed and position loop. In case of resonances caused by high gains, an adaptive notch filter based on Fast Fourier Transform (FFT) is also introduced through online detection of oscillating frequency. Simulation and experiments have been performed verifying the effectiveness of the proposed self-tuning scheme of parameters in AC servo system.