Microfluidics experiments provide insights into transport and chemical processes in porous media, yet measuring evolving concentration profiles remains challenging. Here, the authors introduce a physics-based machine learning toolbox that integrates the non-intrusive reduced basis method, U-Net, and Convolutional Autoencoder to efficiently predict concentration profiles, enabling real-time analysis and tuning of experiments on the fly.
- Ryan Santoso
- Yuankai Yang
- Jenna Poonoosamy