Research On Process Model Based Iterative Learning Control Algorithms

Iterative learning control (ILC) has made great progresses in both theory and practical applications after its first appearence thirty years ago. It has become one of the active research topics in the intelligence control and has attracted great attentions in the control communities. Iterative learning control has its advantages in solving the problems with uncertainties caused by nonlinearities of the plant or modeling errors, because the unknown information can be learned online during the previous system running to offset the absent acknowledgement in the learning process continuously, and improve the performances of the system in a step by step manner. ILC technique appeared originally in the open-loop structures, and considerable progress has been made in theory, however many defects have been found in its practical applications. Therefore model based ILC algorithms were introduced to improve the performance of the ILC systems. The development model based ILC techniques has significant impact on both intelligent control theory and real-time applications.1. The generalization and summary of the control algorithms and theoretical studies on iterative learning control are presented, and the existing defects and research orientations of ILC are pointed. One type of ILC algorithm, based on the minimum variance one-step predictive model for a category of repetitive trajectory process type of CRAMA of discrete-time system, is proposed, such that the output can follow well the reference trajectory, and the ILC has a prediction on control performance;2. For a certain class of plants with CRAMA model, an adaptive iterative learning control (AILC) based on the GPC is proposed, on repetitive trajectory tracking process within a special period of time. Based on the previous process data, the model parameters are identified, and process model data are used to update the control parameters of the ILC controllers. The current input and the deviation of the input from the previous working cycle are computed, and the convergence and the stability of the algorithm are analyzed theoretically;3. For the repetitive processes with unknown state and output disturbances, a model predictive control algorithm with iterative learning compensation is proposed. Based on typical model predictive control, the previous predictive errors of the model are used to compensate the model interference of the system. The control quality of repetitive processes is enhanced by reducing the unknown state interrupt's influence on the predictive models.4. An improved Smith compensation model reference control algorithm for a class of delay integral process is proposed; For linear and nonlinear systems with periodic or aperiodic uncertain interrupts, model reference iterative learning controls (LMR-ILC, NMR-ILC) are proposed, and the state variables of the plant can trace well the state variables of the reference model, on the condition that the plant satisfies certain assumptions and the relevant learning control can be adopted, and it is not necessary that the two models have the same structures and parameters. Convergeny of the algorithm is shown by the use of bound norm;5. For the stabilization problems in discrete-time nonlinear Markovian systems with both available and unavailable status of the model signals, a Bernoulli Iterative learning control (B-ILC) is proposed. A Bernoulli random variable is introduced to express whether the model signals of the system are known, and the algorithm works well for systems with loss of data problems.From theoretical aspect, ILC theory of CARMA models and disturbance model compensation are analyzed in the thesis, respectively for a class of model reference ILC theory of linear and nonlinear systems, the stabilization problems of several types of plants are tackled, and promising conclusions are achieved.From algorithm aspects, several algorithms such as the adaptive ILC algorithm based on the GPC, the ILC algorithm based on minimum variance prediction, the disturbance compensation ILC algorithm, and a class of model reference ILC algorithms for linear and nonlinear system, and an ILC algorithm nonlinear Markov of Bernoulli random variable are proposed, the stability is analyzed for the proposed algorithms.From application aspects, an operation intermittent polymerization device of the typical chemical unit and PH neutralization reaction process are designed with distenbance compensation of ILC algorithms, the simulation experiment are satisfactory.