Minimisation of batch run lengths using predictive control and statistical models

Batch processes are gaining ever increasing importance in manufacturing industries. They are particularly prevalent in the polymer, pharmaceutical and specialty chemical industries where the focus is on the production of low-volume, high-value added products. Yet, while advanced control of continuous processes has progressed significantly over the last few decades, the characteristics associated with batch processes make them particularly challenging to control. These include presence of nonlinear and time-varying dynamics, lack of on-line sensors for product quality variables, frequent operation close to process constraints and an abundance of unmeasured disturbances.In batch processing the objective for the control system can be divided into Batch End /Point Control and Trajectory Tracking Control problems. The fundamental difference between these two types of control problems is that an end-point controller is concerned with ensuring that the quality of the product at the end of a batch meets target specifications, whilst trajectory tracking involves the regulation of product quality to a, typically, time-varying set-point as a batch progresses. Another highly relevant control problem that has not yet been effectively addressed by the academic community is the reduction of batch run length. In fact, the ability to reduce batch run length, while also ensuring that the final product conforms to stringent quality specifications, is arguably the most critical business driver in batch processing industries. The aim of the proposed project is to develop a novel Model Predictive Controller that is capable of addressing a critical operational objective in industrial batch processing, which is real-time reduction of the batch run length. The MPC controller will employ a multivariate statistical data-driven prediction model and will also be applicable to both trajectory tracking and batch end-point control problems for processes that exhibit variable batch run lengths and contain irregular measurements of the controlled variables.The novelty of the proposed project stems from the fact that none of the existing advanced control techniques provide solutions to both the trajectory tracking and batch end-point control while dealing with variable batch run lengths and irregular measurements of the controlled variables. Also, none of the existing controllers address the critical control problem of batch run length minimisation. In contrast, the controllers developed in the proposed project will address all three control problems (trajectory tracking, batch end-point control and batch run length control) while also tolerating the presence of variable batch run lengths and irregular measurements of the controlled variables.