GEEK HUB
Simulating a Printed Heat Exchanger
By Nigel Ravenhill
A solution to rising home energy costs in Europe could well be a combination of curiosity, patience and Siemens’ portfolio of design and analysis software. At least for Siemens Engineer Matthias Duerr.
During the past 18 months, the manufacturing technologies and automation expert has been prototyping a heat exchanger design for his house. It’s been a trial-and-error education as he has progressed from Computational Fluid Dynamics (CFD) newbie to increasingly confident and competent non-expert.
Duerr’s self-made challenge was to improve the energy efficiency of his 1970s home without incurring major expense or necessitating extensive remodeling. Installing a central regenerative ventilation system—while desirable—would require far too much renovation work and invasive piping. To meet his goals of relatively low cost and minimal structural impact, Duerr identified small room-sized recuperative ventilation as the technology area he needed to study and pursue.
An early heat exchanger (left-to-right): Cross section. In-operation. What it really looks like
Trial #1: Early Failures
Duerr’s initial prototypes were both failures and lessons. Built in a basement lab he had outfitted with 3D printers, thermal imaging and other gadgets, these counter-flow heat exchangers proved too bulky and ineffective.
“I found that I had fallen into the same trap as I had on previous occasions, too focused on designing for 3D printing and not freeing my mind of the limitations of traditional manufacturing, such as common extruded plastic profiles where the manufacturing process does not allow for undercuts and changing cross sections. 3D printing largely doesn't have to deal with them. The design approach could follow a much more physics requirement rather than be limited to tubing catalog options from suppliers that utilize traditional manufacturing processes.”
Lessons from University
Reflecting on aerodynamics lectures from his university days, and later gas turbine projects as a professional, Duerr recalled that heat exchanger efficiency improves as wall thickness decreases and a certain type and amount of turbulence is brought in. To avoid creating a laminary flow in which heat exchange is limited to the outside layer (or lamina), his design needed to spin and rotate air while varying the cross section.
Trial #2:
Most 3D printing software used for the slicing of designs provides a mode called “Spiral Vase”. Using “Spiral Vase” you can quickly print a single-layer surface (approx.0.4mm thick). Shapes that gradually change (like aerodynamic designs) can be very accurately rendered. Duerr used Siemens´ Solid Edge CAD software to define a single body, which, despite having only a single cross section, includes a high surface and several channels to guide the flow.
Once printed, the Spiral Vase design was inserted into its airflow container, a PVC wastewater pipe connected to two bilge blowers. Several thermometers were attached to measure temperature while an anemometer measured air speed. Duerr crossed his fingers and let the air flow.
“I was both disappointed yet satisfied. The incoming air, traveling about 5m/s, rose 5°C while passing through the heat exchanger. Better than opening the window but not very impressive. As my first real attempt, though, I was satisfied and immediately began to think about improvements. How could I modify the exchanger pattern? What about adding more and smaller quasi-channels or improving distribution over the cross-section area? How would more length and twisting, or better conducting material and outer insulation help? I had a lot of questions to answer.”
As Spring approached, Project “Heat and Cool My House for Less” went into hibernation as Duerr took a break.
Spiral Vase design for a heat exchanger surface
An Education in CFD
Duerr’s paused heat exchange project restarted when he decided to explore 3D flow simulation through Siemens’ Simcenter™ STAR-CCM+™ CFD tool, part of the company’s broad portfolio of simulation software covering applications as diverse as CFD, computational chemistry, electromagnetics, automation and process workflows.
Like any new software user enthusiastic about progress, he started with tutorials. Several tutorials into his CFD apprenticeship, though, he already understood the clear difference between CAD and CFD software; CAD programs are used to design tangible objects, whereas CFD tools are for designing fluid bodies and describing their boundaries, properties and relations.
CAD data can describe the outer shape of fluid bodies, but surface meshes are what really matter when working with fluid bodies. While programs like Solid Edge can easily convert CAD data into surface meshes, how you work with these meshes is critical—and what ultimately advanced Duerr’s heat exchange simulation.
The body imported from CAD is split into several surfaces, which are then used to assemble the fluid bodies, parting walls, fins, etc. In Duerr’s model above, the hollow shell of the heat exchanger is a perfectly complete part of a CAD model. In his CFD context, though, “end caps” (inlets and outlets) had to be modeled to fully close the fluid space.
The body defined by the surface meshes resembles what is known within the 3D-printing community from STL- and 3MF-data, also involving criteria such as watertightness, and surface orientation. There are key differences: the surface mesh in 3D printing is not required after slicing, whereas the CFD structure of the surface mesh influences the generation of the fluid body’s volume mesh and thus energy and matter exchanges between adjoining bodies. Consequently, there are requirements for the congruency of adjoining surface meshes to align structure of the surface meshes that are addressed by the imprint function.
After disassembling the CAD data into surface meshes, re-assembling them and adding end caps, the result was two adjoining “regions” which circumscribe the respective fluid volumes:
- Outward Flow: The stream of initially warm exhaust air within the exchanger surface.
- Inward flow: The stream of initially cold fresh air between the exchanger surface and outer cylindrical wall.
3D printed heat exchanger
Test in operating environment
Editor for meshes (with exchanger surface and “endcap”) and a close-up of a complicated junction
Meshed Heat Exchanger (Cylinder shown); detail showing 2 types of mesh and 3 boundary layers
Meshed Heat Exchanger (Cylinder not shown, inner endcap (Inlet) not shown)
Early velocity view of the inner flow
Left: Cold from the outer inlet reducing down the cylinder’s skin temperature, while the pressure begins to move the fluid. Right: No exchange of heat. Oops!
How Does Simulation Happen?
Essentially, you have to fill the “regions” with a static volume mesh consisting of cells. Calculations for energy and matter exchanges between cells, etc. are then performed for each cell in each simulation cycle. To advance from describing geometry to this point, Duerr needed to provide more input, specifying different criteria such as orifice type, physical behaviors and properties.
He later learned from a much more knowledgeable colleague that his somewhat casual selection of inlets and outlets without specifying velocity only made another somewhat unusual design decision work: the use of a single “continuum” for both inward and outward flow. A more experienced CFD enthusiast would have chosen two continua (specifying what’s in the fluid volumes) respectively, allowing for opposing initial flow velocities.
The volume mesh, however, was still missing. Decisions about the appropriate level of detail of the simulation needed to be made because the physical effects in different parts of a flow vary; life is much more turbulent for the fluid at the boundary between a fluid and a solid than in the middle of a laminary flow, for example.
In highly turbulent areas, more (read: smaller) and better adapted cells are needed. While the shape and size of the cells are not physics properties, they affect the accuracy of the simulation to the prevailing physical effects. Hence, there are several types of mesh generators to choose from, manually, if the automatic selection needs to be overridden by an expert. Finally, Duerr had to describe the applicable physics models because there’s a clear difference between a supersonic flow and a gentle breeze (much closer to his heat exchanger speed).
Trial #3:
The art of creating a useful simulation is to abstract and simplify the problem to such an extent that real-world hardware can generate a meaningful result reasonably quickly, say, in the time required to make an espresso compared to a 3-day ski weekend in Garmisch Partenkirchen.
Much like a sculptor chooses what to remove and what to leave, the CFD aspirant must decide what to select, approximate and ignore. How did Duerr then cope, not being an aerodynamics engineer?
“Quite frankly, I began by copying the settings from a very basic tutorial—not an optimal solution but it permitted me to create my very first working model. Based on this, I started to modify the physics models and the mesh, playing with different turbulence models, boundary layer thickness and surface roughness, whilst bearing in mind the indicative measurements from the test setup. I was able to see where differences occurred (and what stretched my workstation to its processing breaking point). After some more minor preparations, including defining velocity, speed, temperature, I started the whole thing and watched.”
The Necessity of Describing Your Work
It wasn’t long before Duerr identified an obvious error; failure to describe the heat transfer. The two volumina were ignoring each other. The error was a “bad setting” caused by only specifying an aerodynamic property so there was no heat loss or gain on the heat exchanger surface. The correction was to identify the surface as a “baffle”, fulfilling its role as an interface.
Progress Through Iteration
As Duerr’s CFD knowledge has grown, he has continued to test other parameters and introduce more variables such as decreasing volume flow, raising the temperature difference or replacing the material from plastic to aluminum. As a CFD professional, this is the point where the benefits (cost-savings, etc.) from avoiding making lots of physical prototypes would become patently obvious, including running multiple “what if?” scenarios.
After 18 months, tutorials, clarifying questions to colleagues and many simulations, Duerr has become a serious CFD amateur. Siemens’ Simcenter STAR-CCM+ has enabled him to skip building physical prototypes while testing dozens of configurations. In his pursuit of a cozier house in winter and a cooler home in summer, Duerr’s simulation journey has shown that learning never stops for a curious engineer.