A 2D-FM model-based robust iterative learning model predictive control for batch processes

Latent variable iterative learning model predictive control for multivariable control of batch processes

Iterative learning control (ILC) is often used to eliminate repetitive disturbances and improve the tracking performance in the industrial processes. Due to its disability of reacting against real-time disturbances, the integration of model predictive control (MPC) and ILC constitutes the model predictive iterative learning control (MPILC) to improve its ability of rejecting non-repetitive disturbances. MPILC is suitable for linear models which has poor performance for highly nonlinear systems. A nonlinear MPILC (NMPILC) global controller based on T-S model is put forward in this paper. The global controller is composed of a set of local MPILC controllers designed applying the local linear models of the T-S model. The convergence property has been demonstrated. A PWR nuclear power plant is adopted to illustrate the performance of the NMPILC.

In this paper, an integrated tube robust iterative learning model predictive control (Tube‐RILMPC) strategy based on the dynamic partial least squares (DyPLS) identified algorithm is proposed. The problems of large amount of online data calculation and input and output variable dimensions disaster in the original variable space are solved. This integrated Tube‐RILMPC strategy enhances the tracking performance and robustness of two‐dimensional (2D) control system. The output trajectories of system are located in the tube the nominal trajectory, which reduces the modelling error caused by the uncertain model. Based on the worst‐case performance index of ellipsoidal uncertainty and polytopic uncertainty, a robust iterative learning control (ILC) strategy is designed. Finally, the superiority of the proposed control algorithm is verified by comparative simulation.

Unlike continuous processes, a batch process has a certain period of operation time, and there are a number of batches in a typical operation. Hence variables in a batch process have dynamics in two dimensions, along time and across batches. Besides, batch processes involve large transient phases covering a wide range of operating envelopes, which cause challenges in both modeling and control. To meet the control objectives of batch processes, set-point tracking and disturbance rejection, iterative learning control (ILC) has been widely attempted. This thesis is concerned with the optimal design and performance assessment of ILC based on the minimum variance benchmark. When performance of ILC is unsatisfactory, alternative control strategies should be considered. Generalized predictive control (GPC) is a popular control strategy for continuous processes. Developing a two-dimensional GPC structure for batch processes is another focus of this research. Finally, ILC and suggested GPC are compared through simulation studies.

Successive Linearization based Stochastic Model Predictive Control for batch processes described by DAEs

Batch/semi-batch quality control has attracted many attentions among the chemical engineering researchers recently. In this work, a novel empirical model based quality control approach for batch processes is developed. The model implemented is based on independent nonlinear principal components obtained by principal surface analysis-a very hot topic in the area of statistical modeling. The partial initial conditions, intermediate measurements obtained from each batch are analyzed using this approach to extract the independent nonlinear principal components. An empirical model with the nonlinear principal components as inputs is built to predict product quality. Correcting actions based on the initial conditions and intermediate measurements are taken based on the predictions of the empirical model. In this work, the problem formulation and solution method are presented. The illustrative examples include a simulation example and experimental study of batch reactive distillation process. The simulation and experimental results show the proposed approach is highly promising.

This paper summarizes recent progress in the area of estimation and control of batch processes. The task of designing effective strategies for the estimation of unmeasured variables and for the control of the important outputs of the process is linked to our need to optimize the process and its success is depended upon the availability of a process model. For this reason we will provide a substantial focus on the modeling issues that relate to batch processes. In particular we will focus attention on the approach developed in our group and referred to as “tendency modeling” that can be used for the estimation, optimization and control of batch processes. Several batch reactor example processes will be detailed to illustrate the applicability of the general approach. These relate to organic synthesis reactors and bioreactors. The point that distinguishes tendency modeling from other modeling approaches is that the developed Tendency Models are multivariable, nonlinear, and aim to incorporate all the available fundamental information about the process through the use of material and energy balances. These models are not frozen in time as they are allowed to evolve. Because they are not perfectly accurate they are used in the optimization, estimation and control of the process on a tentative basis as they are updated either between batches or more frequently. This iterative or adaptive modeling strategy also influences the controller design. The controller performance requirements and thus the need of a more accurate model increase as successive optimization steps guide the process operation near its constraints.

A novel combination of iterative learning control (ILC) and model predictive control (MPC), referred to here as model predictive iterative learning control (MPILC), is proposed for glycemic control in type 1 diabetes mellitus. MPILC exploits two key factors: frequent glucose readings made possible by continuous glucose monitoring technology; and the repetitive nature of glucose-meal-insulin dynamics with a 24-h cycle. The proposed algorithm can learn from an individual's lifestyle, allowing the control performance to be improved from day to day. After less than 10 days, the blood glucose concentrations can be kept within a range of 90-170 mg/dL. Generally, control performance under MPILC is better than that under MPC. The proposed methodology is robust to random variations in meal timings within +/-60 min or meal amounts within +/-75% of the nominal value, which validates MPILC's superior robustness compared to run-to-run control. Moreover, to further improve the algorithm's robustness, an automatic scheme for setpoint update that ensures safe convergence is proposed. Furthermore, the proposed method does not require user intervention; hence, the algorithm should be of particular interest for glycemic control in children and adolescents.

Design and Analysis of Integrated Predictive Iterative Learning Control for Batch Process Based on Two-dimensional System Theory

Review for "High‐order iterative learning model predictive control for batch chemical processes"

Decision letter for "High‐order iterative learning model predictive control for batch chemical processes"

On-line batch process monitoring using a consecutively updated multiway principal component analysis model

In this paper, an iterative learning model predictive control algorithm based on state -space model is studied for multiple input multiple output systems with pure delay, and the effect of the iterative control law is analysed. The main contributions of this paper are as follows: first, it solves the inverse problem of unsatisfactory rank caused by pure delay in conventional predictive control, and discusses the solving conditions of iterative learning model predictive control algorithm under the state-space model. Second, it adds the error forward difference term to the iterative control law of iterative learning model predictive control algorithm, which improves the ability of suppressing noise interference. Simulation results show that the proposed method is effective, and it is shown that the suppression effect of noise interference can reach more than 40% in the case of pure delay in the system.

The 2D state space system model is established with state delay and actuator faults, and what is more, the iterative learning control (ILC) law required by the batch process model is designed. Using the knowledge of 2D system theory and an iterative learning control law, the 2D state space model can be changed into an equivalent 2D-Roesser closed-loop system model. According to the optimized cost function and Lyapunov stability theory, sufficient conditions for the solvability of model predictive control (MPC) problems in the form of linear matrix inequalities (LMIs) are given. A new solution is proposed which depends on the bound of the state delay. At last, the effectiveness of the method is proved through simulation experiments.

For batch processes with small time delays and actuator partial fault, the existing methods based on iterative learning control still have some limitations, including the conservative and computationally burdensome of stability conditions and the limited fault-tolerant control capabilities. For this background, an iterative learning robust predictive fault-tolerant control method is developed, which integrates the Lyapunov–Razumikhin function method and derives stability conditions based on robust positive definite invariant set and terminal constraint set. With small time delays, the stability conditions of the system deduced using the Lyapunov–Razumikhin function are solved at a lower computational cost, which is due to the fact that the dimensionality of the stabilization condition is directly related to the size of the time delay, and thus the small time delay implies a lower dimensionality. Especially, the computational effort for solving the stability conditions online is reduced, allowing real-time control law gains to be obtained and combined with historical batches of control inputs, reducing the learning cycles of the system, and realizing stable tracking of the setpoints within shorter operating batches. Moreover, the robust positive invariant set and the set of terminal constraints are able to constrain the state of the system within a safe range for all possible uncertainties, bounded disturbances, and faults. This makes the proposed methods based on them more robust and fault tolerant. Finally, a nonlinear batch reactor is used as an example to demonstrate the effectiveness and feasibility of the developed method.