Robust Predictive Control Design And Batch Optimization For Time Delay Response Processes

In this thesis,robust control design and batch optimization methods are developed for time delay response processes widely operated in industry,based on dead-time compensator(DTC)and a two-degree-of-freedom(2DOF)control structure.Firstly,based on the undelayed output prediction and low-order transfer function models often used in discrete-time domain,a 2DOF control design method is proposed for stable processes with time delay.The plant time delay can be equivalently taken out of the control loop based on DTC and the nominal plant model,such that the controllers can be analytically designed in terms of the delay-free closed-loop.The disturbance rejection controller is inversely derived by proposing a desired closed-loop transfer function.The optimal set-point tracking controller is designed in terms of the internal model control(IMC)theory.There is a single adjustable parameter in each controller,which can be monotonically tuned to make a good trade-off between the nominal control performance and robust stability.Meanwhile,the control performance is quantitatively analyzed in terms of the integral absolute error(IAE)specification for the set-point tracking and load disturbance response.Then,to simplify the control structure and suppress measurement noise,a low-pass filter is introduced with constraints to construct the undelayed output predictor,such that a simplified predictive 2DOF control method is proposed for stable,integrating and unstable processes with time delay.The negative impact from integrating,unstable and slow dynamic poles in such transfer function models is eliminated by using this filter,such that the control performance can be significantly improved.Analytical controller design methods are given in the frequency domain and discrete-time domain,respectively.Illustrative examples from the literature and temperature control experiments of a 4-liter jacketed reactor demonstrate the effectiveness and advantage of the proposed method.Based on the above simplified delay-free output predictor,a control scheme is proposed to combine the classical PI with iterative learning control(ILC)for batch processes with time delay.Based on the measurable output information rather than the internal state information,the PI controller in the inner loop is derived in terms of the IMC theory.Based on the two-dimensional(2D)historical information and a 2D delay-free Roesser model,a sufficient condition for guaranteeing asymptotic stability of the time delay system is established.The iterative learning gain matrices in the outer loop are derived by solving the corresponding linear matrix inequalities,which can ensure the convergence of tracking error along time and batch directions.An illustrative example and iterative learning temperature control experiments are used to demonstrate the superiority of the proposed control scheme.