The present work pursues the ability to increase the fundamental understanding of and to improve numerical modelling capabilities for Compact Heat Exchangers (CHE). This goal is achieved by numerically and experimentally studying the small scales of flow, then in- corporating these results into a novel reduced order model (ROM) of a full heat exchanger. The result is a numerically efficient modelling approach that significantly improves numer- ical modelling of CHE’s.

The heat transfer enhancement surface of focus, the turbulizer, is studied in great de- tail with high resolution Computational Fluid Dynamics (CFD), using experimental flow visualisation and Laser Doppler Velocimetry (LDV) measurements to validate the results. The modelling process explores a variety of turbulence models and simulation methodolo- gies, finding that a Large Eddy Simulation (LES) model with several modifications to the turbulizer geometry to replicate manufacturing process pressure drop predictions within 7% of experimental results and heat transfer within 15% of experimental results. Excellent correlation is also observed with predictions of transition to unsteady flow. Flow visualisa- tion provides excellent correlation with predicted flow patterns at low Re . The validated turbulizer model is used to investigate flow conditions through a wide range of Re and flow incidence angles, which have not been previously studied.

Construction of the reduced order model leverages data obtained during the detailed simulation of the turbulizer under a variety of flow conditions, mapping the appropriate Nu and f_D to a porous media heat transfer framework. This framework is used to enforce the heat transfer and pressure drop calculated based on the detailed modelling phase. Model lookup performance is investigated using both an artificial neural network (ANN) and bi-linear interpolation. The ANN approach provides the best overall performance. Implementation of the ROM and turbulizer flow data is undertaken within the frame- work of STAR-CCM+, using field functions and user defined code to interact with the proposed model. Heat transfer is validated against experimental test results of a heat exchanger design which has previously been problematic for analytical models to accom- modate. The results indicate an approximate halving of the error in pressure drop and heat transfer predictions made by numerical and analytical models, respectively. This in- dicates that the proposed novel ROM methodology provides a significant increase in the numerical predictive capabilities of complex heat exchanger models under a wide variety of flow conditions.