Edgar Romero

Empirical investigation of heat transfer in additively manufactured heat exchangers

In this PhD, Edgar will be focussing on developing a methodology that enables the user to rapidly and iteratively design a heat exchanger core hat meets a set of heat transfer and pressure drop requirements whilst adhering to spatial constraints.

Metal additive manufacturing (AM) is viewed as a key enabling technology for the next generation of thermal management solutions (e.g. heat exchangers). Heat exchangers, used to transfer heat between two fluids, are essential components in many engineering systems in sectors such as aerospace, automotive and energy. The harmony between AM and heat exchangers arises through the relative ease with which complex and intricate internal geometries (channels) can be produced without the need for costly fabrication stages. As such, AM heat exchangers have already established themselves as highly performant, compact and lightweight alternative to traditional heat exchanger concepts.

However, AM presents significant challenges in terms of development costs and time, particularly where iterative production might be expected. A typical, single machine facility is likely to cost in the range of £1 million, and titanium powder feedstock costs approximately £400 per kg. A heat exchanger with dimensions of 200 by 200 by 200mm would take approximately 10 days to produce. As such, to iteratively develop a new heat exchanger concept using this technology would easily exceed the £100k mark in terms of development cost.

Cellular geometries and, particularly, triply periodic minimal surfaces, have gained a lot of attention in both the literature and industry within the context of heat exchangers. These mathematically-defined geometries present various appealing properties. Most importantly, these are all cellular in nature and split the cell into two equal or unequal volumes, which remain interconnected between adjacent cells. Therefore, two different fluids can travel through the two networks and always be in close proximity (separated by a thin wall) without physically mixing.

Several challenges exist in this field, however. While these are mathematical designs and, hence, can be modified in limitless ways, they do not inherently maximise heat transfer and/or minimise pressure drop, which are the two main challenges in heat exchanger design. The authors believe that there is plenty of scope for at least some of these geometries to be altered to find more optimal designs than the default minimal surfaces as provided by the conventional equations. A part of this project involves comparing different designs of multiple minimal surfaces to provide insight into which of and how these minimal surfaces could be made optimal for specific heat transfer, pressure drop and mass constraints.

Another related but worth-highlighting gap in the literature is the overuse of CFD without sufficient experimental evidence. Most of the work involving additively manufactured heat exchangers is either industry-lead, and therefore fairly vague and opaque, or, in the opinion of the authors, academically-lead but not rigorously validated. The reason for our scepticism lies in the fact that several review papers highlight the difficulty in accurately predicting performance with CFD (which usually underestimates that pressure drop significantly) and the lack of understanding of the effect of surface roughness.

It is the intention of the authors to experimentally investigate the heat transfer and pressure drop in either all or some of the selected  minimal surfaces to enhance the knowledge base of flow patterns and behaviours in these intricate and complex geometries, characterised by varying levels of flow mixing, recirculation and turbulence. These depend on the working flow regime, which is expected to be between laminar and transitional for air.

During this campaign, the emphasis will be on surface roughness. As mentioned previously, a lot of resources are invested in the production of heat exchangers with additive manufacturing. Furthermore, two equal designs can vary in performance significantly if they are manufactured with different machines or machine settings. If surface roughness can be isolated successfully from the theory, data obtained from designs with smooth surfaces (like SLA or even within CFD) could be used to predict performance in metal-based prototypes, which take much more effort to produce.

It is expected that this research will play a role in the current trends of significant reduction in emissions in the aforementioned industries, owing to a reduced mass and therefore energy/fuel savings. In addition, enhanced performance will also help to recover and harness wasted heat within these systems. Looking further, it is thought that this could help make future aircraft propulsion and power generation systems viable, such as hydrogen fuel cells and widespread electrification.

The aim of this project is to develop a methodology that enables the user to rapidly and iteratively design a heat exchanger core that meets a set of heat transfer and pressure drop requirements whilst adhering to spatial constraints. The current vision is to combine novel heat transfer modelling with an algorithmic design approach. This will be used to automate the design of the core geometry and therefore reduce the engineering overhead and reduce the time required to reach a new proposition. A heat exchanger test bed will be designed and built to facilitate the thermofluid characterisation of the specimens as well as serve as an integral part of the design methodology. validate the modelling work but also to form an integral part of the design methodology by using it as hardware-in-the-loop.

While lead time and development costs are a major priority in this project, the trade-off with high performance and efficiency must be managed successfully. Due to its unparalleled versatility when producing geometries, AM allows for highly efficient geometries that, while more expensive to produce than traditional heat exchanger designs, could noticeably reduce the running costs of the systems to which they will pertain. Therefore, the ideal version of this methodology would also enable engineers to produce an optimal heat exchanger design for any given application, although this will likely be limited to a parametric optimisation algorithm – an optimal version of a specific heat exchanger concept. A semi-empirical methodology is thought by the researchers to be the most appropriate way of achieving the desired accuracy of thermal modelling and, thus, minimise the probability of expensive and time-consuming alterations to the heat exchanger design (or even a complete design overhaul).