Machine-learning-assisted population balance modelling for pharmaceutical crystallisation process design

Project Vision

The development of active pharmaceutical ingredients (APIs) increasingly demands flexible, predictive, and resource-efficient modeling tools to support sustainable, scale-up, and uncertainty-conscious process design. While classical Population Balance Models (PBMs) provide deep physical interpretability and extrapolation power, they are often limited by parameter uncertainty and high experimental costs. Conversely, purely data-driven machine learning models offer fast predictions but lack physical consistency outside of collected experimental data ranges.

Our research group bridges this gap. We build Machine-Learning-Assisted PBM Frameworks that synergistically integrate first-principles chemical engineering models with state-of-the-art machine learning tools to deliver computationally efficient, robust, and highly accurate digital twins for crystallisers.

Project Duration

2026-2031

Team

• PI: Kensaku Matsunami