Study On Tuning Strategies Of Fractional Order Controllers For AC Servo Drive System

High speed and high precision applications demand for a high control performance of AC servo drive system. To obtain a satisfactory control performance, advanced control techniques are needed to guarantee both dynamic performance and robustness of the servo system. Traditional PID control method has the advantages of simple structure, high efficiency and easy to implementation. However, since the nonlinearities and uncertainties always exist in the servo drive system, it will influence the control performance of traditional PID control method. Therefore, this paper will adopt the fractional calculus to the traditional PID control technique. Fractional order PID (FOPID) control method will be studied. Meanwhile, this paper will focus on the controller tuning method of FOPID and seek for stable, efficient and practical FOPID tuning strategies.For the purpose of improving the control performance of servo drive system, combined with the National Natural Science Foundation of Chine, this paper will focus on the study of FOPID control method and its tuning strategies. Three classes of tuning methods will be studied:graphical tuning, numerical tuning and online tuning. The detail contents are as follows:A thorough analysis will be conducted on the mathematical model of the speed loop of servo drive system. A fractional model will be constructed and an optimization based fractional model identification method will be proposed. Theory and experiments all prove that fracitonal model can better describe the AC servo drive system.Both model based and model free graphical tuning methods are studied. The purpose of the graphical tuning method is to compute the stabilizing and H? region of FOPID controllers. For model based tuning, an interval model based graphical tuning method will be proposed. Applying this method, the tuned controller can have a strong robustness to the model parameters uncertainties. For model free tuning, a frequency response based graphical tuning method is proposed. This method can be applied for both linear and nonlinear FOPID-type controllers.For numerical tuning, both model based and model free methods will be studied. For model based tuning, an enhanced stochastic multi-parameters divergence-based optimization (enhanced SMDO) based tuning method will be proposed. This method considers both time and frequency domain specifications of the servo system. It can be applied for both integer and fractional order system. The traditional SMDO method is enhanced by improving its convergence speed and accuracy. For model free numerical tuning, a virtual reference feedback tuning (VRFT) based FOPID controller tuning method is proposed. The proposed method is extended for its online version.For online tuning of FOPID controllers, both model based and model free method are researched. A model based online tuning method is proposed. The method is based on recursive least square (RLS) and controller parameters database technology. This method adopts RLS to online identify the model parameters and searches the optimal parameters in the database. The method has a low computational burden and easy to implement. For model free online tuning, a wavelet neural network (WNN) based tuning method is proposed. The method uses WNN to replace the traditional BP neural network to improve the adaptability and fault tolerance capability. Meanwhile, the offline tuned parameters are adopted to refine the online tuning. Therefore, the servo system can always maintain a good control performance during the initial and later phase. The proposed method can be used for both linear and nonlinear FOPID-type controllers.Both linear and rotary servo platform are built. Hardware and software design are conducted and proposed algorithms are implemented on the platforms. Experiments verified the validity and accuracy of the proposed methods.