Skip to main content
Research Project

AI and machine learning for process modelling

Combining data-driven and hybrid models with first-principles and CFD modelling for chemical processes, and teaching the methods through short courses and collaboration.
Share this project
Post to LinkedIn
Post to X
Share on Facebook
Send via WhatsApp
Share on Bluesky
Share via Email
AI and machine learning for process modelling

Why it matters

Process industries are under pressure to cut energy use and waste while staying competitive. Plants produce large amounts of data, and detailed simulations such as CFD can describe equipment well, but each alone falls short. Data-driven models struggle with noise and with conditions not seen before. First-principles and CFD models are accurate but often too slow for design iteration or online use.

The problem

  • Process data are noisy, correlated and heterogeneous.
  • High-fidelity simulations are expensive to run repeatedly.
  • Machine learning models must be reliable and safe before they are trusted in a plant.
  • Few engineers are trained in both process modelling and machine learning.

What we do

  1. Data-driven models. Use latent variable methods, tree-based models and neural networks for prediction, monitoring and feature extraction in chemical processes and multiphase flows.
  2. Hybrid models. Combine machine learning with process models and CFD, so that learned models can reduce the cost of simulation while keeping the physics.
  3. Teaching. Develop courses and workshops that bring the two traditions together for chemical engineers.

Approach

The work treats machine learning as a complement to modelling, not a replacement. Our short course From Data to Dynamics: Advanced Process Modelling covers data handling, principal component analysis and partial least squares, tree-based models and neural networks, and CFD for multiphase flow, with hands-on sessions.

We also prepared a SPARC proposal with IIT Kharagpur and NIT Calicut on AI-driven modelling for process industries, covering AI-assisted high-fidelity modelling, knowledge exchange and training. Outlines for an introductory AI and machine learning course and a monograph on AI in process engineering were drafted alongside it.

Outputs

Collaborators

IIT Kharagpur, NIT Calicut and Curtin University.

Contact

For collaboration or student projects in AI-assisted process modelling, contact the SMILE lab.