Characteristic Models And Iterative Learning Control Methods Of Pmlsm Servo-Systems

Characteristic modeling theory provides a simple way to build parametric linear models ofthe PMLSM servo-systems. By utilizing the iterative identification methods, characteristic modelbased adaptive iterative learning controller can achieve an accurate tracking performance for thePMLSM servo-systems. Meanwhile, in the presence of random initial state errors, varying-orderstrategy based sampled-data ILC can achieve a better control performance when compared withthe conventional one. This thesis includes the following aspects:1. PMLSM’s mathematical models in the d-q coordinate are derived. Under the "id=0"control strategy, the PMLSM servo-systems, considering the reference speed as systems’ inputand motor’s position as systems’ output, are proved to be the six order linear time-invariantsystems, by using Laplace transform method.2. Based on PMLSM servo-systems’ d-q model and six order linear time-invariant model,systems’ SISO characteristic models and MIMO characteristic models are derived, respectively.Both the theory analysis and numerical simulation show that, if the sampling time is smallenough, then the original system’s output and characteristic-model’s output are equivalent, and allthe parameters are varying slowly along the time axis, which meet the features of characteristicmodel theory.3. A characteristic-model based AILC is designed for the systems, which can be describedby the characteristic models. This controller includes an iterative identification and a regulator.Gradient iteration identification algorithm or forgetting factor least squares iterativeidentification algorithm can be choosed for the identification, and the regulator can be designedthrough using LQ optimal control method or one-step ahead method. Numerical simulations aretaken for the PMLSM servo-systems, and the results show that a good performance can beachieved under the initial state errors.4. A varying-order sampled-data ILC is presented for a class of nonlinear MIMO systemswith perturbed initial conditions, and it is extended to the systems with any relative order. Whenthe system’s initial state errors vary randomly in a certain range, the proposed DIP varying-order strategy has faster convergence speed than DD one, and has higher control precision than theconventional one. The simulation results of the PMLSM servo-systems show effectiveness of theproposed algorithm.