Predictive Control of Batch Crystallization Process Using Machine Learning
Yingzhe Zheng, Zhe Wu
Abstract
This work develops a framework for building machine learning models and machine learning-based predictive control schemes for batch crystallization processes. We consider a seeded fesoterodine fumarate cooling crystallization and dissolution in a batch reactor and present the methodology and implementation of simulation, modeling, and controller design. Specifically, to address the experimental data scarcity problem, we first develop a population balance model (PBM) based on published kinetic parameters to describe the formation of crystals via nucleation, growth, and agglomeration. Then, recurrent neural network (RNN) models are developed using data from extensive simulations of the semi-empirical PBM under various operating conditions to capture the process dynamic behavior. The model predictive control (MPC) scheme using RNN models is developed to optimize the crystallization process in terms of product yield, crystal size, and energy consumption, while accounting for the constraints on the manipulated inputs. Through open- and closed-loop simulations, it is demonstrated that the RNN models well capture the process dynamics, and the RNN-based MPC achieves desired product yield and crystal size with significantly improved computational efficiency.