Polynomial Chaos Expansion (PCE) Based Surrogate Modeling and Optimization for Batch Crystallization Processes

摘要

The paper presents a computationally efficient approach to represent a nonlinear data-driven input/output model between the finite-time control trajectories and the quality index at the end of the batch, based on the approximate representation of the full process model via polynomial chaos expansion (PCE). A batch cooling crystallization system of Paracetamol in water was used to estimate the dependence of output mean length of crystals at the end of the batch on the temperature trajectory applied during the crystallization. The surrogate model was then validated for its performance. Later, the surrogate model was used to determine the optimal temperature profile needed to maximize the mean length of crystals at the end of the batch. The validation and optimization results prove that the experimental data based PCE can provide a very good approximation of the desired outputs, providing a generally applicable approach for rapid design, control and optimization of batch crystallization systems based on experimental optimization.