This project focused on improving the efficiency and reliability of cooling systems by using artificial intelligence to predict and manage nanoparticle deposition in advanced heat exchangers. Heat exchangers are vital components in high-performance environments such as data centres and AI computing facilities, where precise thermal control is critical. However, one of the main challenges in these systems is the build-up of nanoparticles, which can block flow paths, reduce heat transfer efficiency, and increase operational costs.
To tackle this problem, the project developed a hybrid computational model combining fluid dynamics simulations with advanced machine learning algorithms, including Deep Neural Networks (DNN) and Long Short-Term Memory (LSTM) models. Using data from Python-based simulations, the team examined how factors such as heat flux, nanoparticle concentration, tube geometry, and particle size affect deposition patterns inside hexagonal-tube heat exchangers. The DNN model achieved a predictive accuracy of 97 per cent, making it a highly reliable tool for anticipating fouling before it impacts performance. The findings revealed that higher heat flux and nanoparticle concentrations lead to increased deposition, while optimised tube geometries and flow conditions can significantly reduce accumulation. These insights are particularly relevant to AI and data centre cooling, where liquid and nanofluid-based heat exchangers are increasingly used to manage the intense thermal loads generated by processors and GPUs. By integrating AI-driven predictive models into cooling system design and operation, data centres could pre-empt blockages, extend equipment life, and maintain energy efficiency even under high computational demand. This project demonstrates how combining physics-based modelling with machine learning can transform thermal management, supporting the next generation of sustainable, high-performance data infrastructure.