(303e) Pls Based Iterative Learning Control Scheme for Batch Systems

As the current market is highly competitive, the product life cycle is getting shorter and shorter. Batch processes have occupied an important position in the chemical industry and the past persistent economic growth over the last couple of decades. Operation of the batch process is quite different from that of the continuous process, because the former is characterized by prescribed processing of raw materials into products in finite duration. Given a specified quality of batch processes, products are produced and their characteristics are significantly affected by the operating condition of the manipulated variables. In recent years both industry and academia had a strong interest in the development and application of batch-to-batch (BtB) feedback control in the batch chemistry industries. Extensive studies were done to improve or tune the BtB controller, but the development was limited to single-input single-output processes. The control of multivariable systems is not always an easy task due to its complex and interactive nature. Based on a linear multiple input and multiple output model, a multivariate EWMA (exponentially weighted moving average) controller was developed. A multivariable extension of the dEWMA (double exponentially weighted moving average) controller was presented to analyze the robustness and stability conditions. If the determinant of the process gain matrix was very small, the controlled system would be extremely sensitive to any error in the process model. In the multivariable process, the degree of interaction among the input and output variables is usually of high importance. The multivariable statistical techniques, such as principal component analysis and partial least squares (PLS), have been recently applied to process engineering problems involving system monitoring and diagnosis. They can condense the variance of the process into a very low-dimensional latent subspace. As the method with only a few principal components of PLS could capture most of the characteristics of the system pattern in a multivariable process, it is worth extending PLS models to the BtB controller design to overcome the problems currently encountered in the MIMO RtR controller.

In this paper, a new feedback batch control strategy based on PLS model and dEWMA control for the end-point product quality control is proposed. In the batch-to-batch operation, PLS dEWMA control is done by applying feedback from the final output quality of the batch process. It utilizes the information from the current batch to improve quality for the next batch. Using the resulting structure PLS model makes it possible to extract the strongest relationship between the input and the output variables. It is particularly useful for inherent noise suppression. Without the decoupler design, the non-squared MIMO control system can be decomposed into a multiloop control system by employing precompensators and postcompensators of the PLS model constructed from the input and output loading matrices. Then the conventional dEWMA controller can be separately and directly applied to each SISO control loop to address the model errors gradually reduced from model-plant mismatches and unmeasured disturbances. The major advantage of the proposed method is the simplicity of designing the controllers individually based on the SISO control algorithms. It is not necessary to design the MIMO system based on the whole system variables. The coupling effect problem in the MIMO system can be overcome effectively. A simulation process in this study shows that the proposed method has good performance dealing with the drift rejection without any large offset when compared with the conventional MIMO BtB method.