Robust nonlinear model predictive control of batch processes

Abstract

NMPC explicitly addresses constraints and nonlinearities during

Optimal seed recipe design for crystal size distribution control for batch cooling crystallisation processes

“…In seeded crystallisation, ideally the supersaturation is maintained at the desired constant value throughout the entire batch by the application of properly designed control algorithms (Chung et al, 1999;Nagy and Braatz, 2003;Xie et al, 2001;Zhang and Rohani, 2003;Simon et al, 2009a). Supersaturation, generated by cooling, can be consumed by the growth of seeds added, and hence, it can be kept relatively low throughout the batch if enough seeds are loaded.…”

Data treatment and analysis for on-line dynamic process optimization

“…The state estimations that use a time-varying full matrix Q lead to a better performance than the constant diagonal matrix, as it is shown in Figure 1. Although the linearized approach performance has not been as good as the Monte Carlo approach performance, it can be improved whether the parameter covariance matrix C p is available from parameter estimation [5]. Besides, the CEKF has presented the best performance for the state estimation for all the tuning techniques.…”

“…In [3,4], two systematic approaches are used to calculate Q from the parametric model uncertainties and the accuracy of this techniques are compared favorably with the traditional methods of trial-and-error tuning of EKF. Moreover, the NMPC algorithm proposed by [5] takes parameter uncertainty in account in the state estimation through these systematic approaches. Furthermore, the use of data preprocessing and dynamic data reconciliation techniques can considerably reduce the inaccuracy of process data due to measurement errors, improving the overall performance of the MPC when the data is first reconciled prior to being fed to the controller [6].…”

Nonlinear Model Predictive Control of Batch Processes: An Industrial Case Study

“…Note that the above approach leads to a time-varying, full covariance matrix, which has been shown to provide better estimation performance for batch processes than the classically used constant, diagonal ( Valapil and Georgakis, 2000;Nagy and Braatz, 2003).…”