(12g) State Space Model Predictive Control Using Quality Variables for Batch Polymethyl Methacrylate Production

Polymethyl methacrylate (PMMA) is an important industrial polymer with widespread applications ranging from plastics to electronics. PMMA production occurs via a specific recipe in a batch process in order to ensure the optimal product quality. The key parameter to determining product quality is the molecular weight distribution of PMMA. However, measuring the entire molecular weight is a long, complex procedure, therefore the number and weight average molecular weights are used as outcome measures instead. These quality variables are measured at the end of the batch and, as such, must be controlled tightly.

To achieve good quality control of the PMMA process, it is important to develop a process model capable of handling batch data. Batch processes are unique in that they tend towards the low volume production of high value products, as this allows for poor quality batches to be discarded. Each discarded batch represents a significant loss in revenue, motivating the need for advanced batch control approaches and, fundamentally, the development of accurate quality models. Recent advances in computing technology have led to increased amounts of historical data being available, making data-driven modeling a viable choice for model identification.

Developing an accurate process model is the key to good quality control. This work focuses on quality modeling in batch process data using a missing data subspace algorithm that adapts nonlinear iterative partial least squares (NIPALS) algorithms from both partial least squares (PLS) and principal component analysis (PCA) to build a data driven model. The next consideration is to utilize this model in a state space model predictive controller to get improved batch quality for a polymethyl methacrylate (PMMA) process. Traditional PMMA control is done with a proportional-integral controller and setpoint tracking from optimal batches using output variables to indirectly control temperature. By identifying a model with quality variables this work utilizes MPC to directly control the quality of the batch throughout the process. This MPC is then able to show improved performance compared to traditional control.