AI Driven Open Source Framework for Next Generation Heat Exchangers
This project has generated AI models that significantly improve turbulence modelling when designing heat exchanger cooling systems.
Project aims and approach
This project worked on an open-source curated dataset for modelling turbulence in the design of printed circuit heat exchanger (PCHE) and cold plate cooling systems. These systems are heavily used by many industries tackling electrification and net zero.
To design efficient heat exchangers and cooling systems, accurate computational fluid dynamics (CFD) software including turbulence modelling is needed. Techniques such as large-eddy simulation (LES) and direct numerical simulation (DNS) have high accuracy, but the computing power needed makes them unaffordable for many simulations.
For this reason, Reynolds-averaged Navier-Stokes (RANS) simulations are expected to remain the main tool for predicting flows in engineering over the next few decades, but they have limitations.
This project aimed to train a model based on artificial intelligence which can be used to improve the accuracy of RANS simulations at almost no extra computational cost. The AI model was trained using a variety of DNS and LES data relevant to heat exchangers in different industries.
You can watch a short ‘show and tell’ presentation about this project below, recorded in September 2022. Start from timestamp 18:44.
July 2022 – March 2023
Barriers and achievements
This project has generated AI models that significantly improve turbulence modelling for PCHE cooling systems.
The open-source dataset produced will allow machine learning augmented turbulence modelling to be used immediately in existing computational fluid dynamics tools, and in future developments.
The new features will be incorporated into TOffeeAM’s commercial software products.
The algorithm and dataset have been made openly available via Github. The company expects that the framework and methods it has developed will be extended by the community to cover different configurations and address new challenges around turbulence modelling that will undoubtedly emerge, as industries develop technologies for green transformation.