# Aerodynamics of Propulsion

### Help Support Homebuilt Aircraft & Kit Plane Forum:

##### Well-Known Member
Supporting Member
Yeah, I'm more curious just to see what numbers a modern design with known dimensions gets.

I'm circling the drain of being tempted to DMLS print an integrated heat exchanger for charge air, coolant, and oil, but that's beyond my personal comfort range so I'd end up contracting with someone if I do look into that path... but boy, does it make some awesome things possible.

#### rv6ejguy

##### Well-Known Member
Supporting Member
I updated my last post if you're looking to spend some money on testing...

##### Well-Known Member
Supporting Member
Thanks!

Some inspiration (and some more) for 3D printed heat exchangers. I doubt I'll take this route (it wouldn't be cheap, at our size; the EOS M 400 is bounded by a 16" cube which would be more than enough for block plus fittings, but it's a million dollar machine and service bureau pricing reflects that), but it's compelling to think about. Would allow designing to a chosen k_p, printing in cheap plastic for flow measurements and refining. Complex structures possible, integrating multiple heat exchangers into one block. Could line up the fins/tubes/etc with the local airflow from the diffuser, preventing separation within the block towards the edges. Avoid corners on the block itself, even. So many interesting opportunities. So complex to dig into.

#### rv7charlie

##### Well-Known Member
Supporting Member
'Separation' in the core is a requirement for efficient operation.

##### Well-Known Member
Supporting Member
My understanding is that turbulence (small vortices) in the core is a requirement; separation (large vortices) is not. If a fin fully “stalls” towards the edge because it is too far misaligned from the local flow, the large vortex will obstruct the flow between it and the next fin. This is the purpose of their guide vanes which are super-local adjustment of fin leading edge angle to match local flow.

#### dog

##### Well-Known Member
if you were going to all that trouble than the face of your multi Hx would not be flat
perhaps convex,concave,a combination
but not flat

##### Well-Known Member
Supporting Member
Not completely clear that's true. One face is likely to be planar to reduce manufacturing costs significantly. Assuming that's the aft face, that suggests that parallel flow at the outlet of the exchanger is still optimal. If inflow velocity is uniform (which is the goal of the diffuser), then optimal depth of the exchanger is constant across its area, which combined with parallel outflow and a planar back surface suggests that the front surface will also be pretty darn close to planar.

#### rv7charlie

##### Well-Known Member
Supporting Member
My understanding is that turbulence (small vortices) in the core is a requirement; separation (large vortices) is not.
We agree; I was assuming that you meant turbulence since actual separation inside the core seems well nigh impossible, unless you're talking about something like the space between the fins of adjacent cylinders on an air cooled engine. No properly designed liquid-air HE would have fins spaced widely enough for separation to happen inside the core. However, laminar flow could happen in a poorly designed core (just as bad).

##### Well-Known Member
Supporting Member
Fair enough. My mental model of a fin (as opposed to a flattened tube) in a modern radiator is as a thin flat plate, which is a very small leading edge radius. Towards the walls of a streamline diffuser there's a significant outward flow component, exceeding the 9° or so stall angle of a flat plate. Naïvely, this would cause a separation bubble ("large"-scale vortex, relative to the leading edge radius) right at the leading edge, obstructing flow between this fin and the next. I'm not sure this is a particularly useful mental model, but it is consistent with the experimental value of the "guide vanes" that essentially just bend this leading edge to align it with the local flow.

#### rv7charlie

##### Well-Known Member
Supporting Member
Sounds like you're conflating the diffuser characteristics and core characteristics. Also, no modern core fin is as simple as you describe. The tubes have 'dents and bumps' to keep the liquid flow inside the tube and air outside the tube turbulent, and the fins have 'wrinkles ' to keep the air turbulent all the way through the core.

dog

##### Well-Known Member
Supporting Member
Am I? I'm think I'm talking about the interaction with them, where the non-parallel flow of the diffuser becomes the nominally-parallel flow within the core. Just like any other plate or airfoil in the flow, there's a negative pressure peak associated with turning the flow. This is a very local effect -- the structure of the core past the leading edge shouldn't effect it much, unless I'm missing something.

#### rv6ejguy

##### Well-Known Member
Supporting Member
AFAIK nobody has been able to properly model the flow in a multi row HX. The change in V, Reynolds number, density and temperature as it progresses through the core and the tiny sizes of the elements make any simple model with simple assumptions, inaccurate. People in the industry rely on experimental measurement to arrive at real numbers- Spearco and PWR for instance. You could extrapolate a short distance from there once you catalog characteristics of several similar designs- perhaps, but the further away you are from what you measured, the less accurate the results are likely to be.

I fail to see how you can model or calculate anything without accurate HX pressure drop, effectiveness or even coolant flow rates. You absolutely need accurate empirical data to plug into any calcs or CFD.

For most of us here, we'll be using available OTS HX cores of X depth, perhaps custom cut to Y width and Z height to give us the best compromise for ground and flight cooling with the least possible drag for our application.

Anyone who has the monetary and engineering means to design and create HXs from scratch- power to you and hopefully you share your data with the community.

As for ducts, same thing applies. The HX messes up any analysis of the internal flow of a duct. Simple assumptions and models don't work from what I've seen and you can't model the system as a whole without an accurate HX model. My Boeing friend has run into many problems here.

In this complex field, it's actually easier to build the test rigs and get the real numbers than making guesses and arriving at an answer which will likely have no basis in actual reality.

Any engineer worth his salt should be able to define and design the experimental rigs to quantify HX performance and then progress onto the duct down the line.

How much energy/time/ money will someone expend on this aspect of their Experimental aircraft?

Last edited:

##### Well-Known Member
Supporting Member
It’s definitely easy to go far beyond the point of diminishing returns here, and I doubt I will. But it’s fun to think about. And a very cool think about planning for DMLS is that the design rules are close enough that you can print in plastic at home in ten hours for $3, and do flow tests for both internal and external flow on the plastic version that should be pretty darn close to an aluminum print. Faster iteration and cheaper than buying, testing, and returning; but that last print is a doozy. #### rv6ejguy ##### Well-Known Member Supporting Member It’s definitely easy to go far beyond the point of diminishing returns here, and I doubt I will. But it’s fun to think about. And a very cool think about planning for DMLS is that the design rules are close enough that you can print in plastic at home in ten hours for$3, and do flow tests for both internal and external flow on the plastic version that should be pretty darn close to an aluminum print. Faster iteration and cheaper than buying, testing, and returning; but that last print is a doozy.
The 3D printed model of the HX for flow testing several designs is very intriguing however in the end, you'll probably be selecting an OTS HX anyway so it's actually easier and faster to do all the flow and thermo testing on that, although not cheaper than the plastic piece. I'd be interested to see some pix on how the printed piece turns out. There will be some fine detail required in the typical pierced/ louvered finning most HXs use today. Would be a cool project.

##### Well-Known Member
Supporting Member
I figure if I were to 3D print for testing I’d probably 3D print for real as well. The EOS DMLS printers are ideal for this (like 90% of their demo pieces and press shots are heat exchangers, because it’s so easy to do radically interesting designs), and even though Fathom (the service bureau I usually use) doesn’t have an EOS big enough, they’re out there.

#### Monty

##### Well-Known Member
A few quick comments after looking at what you guys have been up to:

Heat exchangers can share a common inlet, but they need separate ducts/outlets to solve the issues already discussed.

The radiators in the P51 are archaic and terrible. They hadn't really figured out the area ratio for the heat exchanger is set by the relative heat capacities: IE large area for air, small area for water. Modern radiators are MUCH MUCH better (and lighter!!) than anything the warbirds had available.

The flow coefficient of the radiator and radiator effectiveness are related. The pressure drop is also effected by the heat rejection. The problem you always run into with these things: You need data to make the analytical part work.

CFD: To use CFD for this problem, you replace the radiator with a surface. The pressure drop and heat rejection characteristics of the radiator are applied as boundary conditions. You don't try to analyze the radiator. You analyze the ducts and external flow on the aircraft. Again, you need data from the radiator (pressure drop and heat rejection).

It would take quite a test rig to produce the airflows and heat rejection rates we see in our aircraft. I had a plan to mount a test rig on a car, but life happened. You can get cold pressure drop through the core using a leaf blower and manometer. You need a small section of radiator to get the right flow rate per sq. in. Use a pitot static setup to measure speed. This will get rough numbers, and you can neglect the pressure drop due to heat transfer. (gotta start somewhere).

#### rv6ejguy

##### Well-Known Member
Supporting Member
It would take quite a test rig to produce the airflows and heat rejection rates we see in our aircraft. I had a plan to mount a test rig on a car, but life happened. You can get cold pressure drop through the core using a leaf blower and manometer. You need a small section of radiator to get the right flow rate per sq. in. Use a pitot static setup to measure speed. This will get rough numbers, and you can neglect the pressure drop due to heat transfer. (gotta start somewhere).

#### dog

##### Well-Known Member
Sounds like you're conflating the diffuser characteristics and core characteristics. Also, no modern core fin is as simple as you describe. The tubes have 'dents and bumps' to keep the liquid flow inside the tube and air outside the tube turbulent, and the fins have 'wrinkles ' to keep the air turbulent all the way through the core.
this is a very interesting description, using maximum turbulence to increase conduction to dissipate heat

weird to think about as the turbulence/friction is creating heat itself but in this one case* the increased conduction must be orders of magnitude greater

* thinking of mechanisms rather than natural phenomena, something something, ice cream makers, still want to kick anybody mixing food and physics

Last edited by a moderator:

#### wsimpso1

##### Super Moderator
Staff member
this is a very interesting description, using maximum turbulence to increase conduction to dissipate heat

weird to think about as the turbulence/friction is creating heat itself but in this one case* the increased conduction must be orders of magnitude greater

* thinking of mechanisms rather than natural phenomena, something something, ice cream makers, still want to kick anybody mixing food and physics

We we talk about turbulent flow inside a duct or along a cooling vane, we are talking about where dynamic effects are greater than viscous effects, driving flow to separate and tumble. When we compute the Reynolds number - a dimensionless number that is used to represent the ratio of dynamic to viscous effects - for coolant flow through the tubes and for air flow over the tubes, we get small Re, which means the flows tend to be deeply laminar. Most of the heat transfer is then by conduction between the metal tube and a thin layer of air, but conduction of heat from that thin air layer to the rest of the air is pretty lousy.

As to maximizing turbulence, that is tough to do. To raise Re you either need to seriously raise velocities (more drag) or increase length of air flow along the tube. And we are talking a lot of increase to get to turbulence threshold, so that is an undesirable approach. We can put those little louvered fillers between tubes. These do two things: They cause the air flow to turnover (tumble) and they increase the hot area exposed to the cooling air. Turnover is very effective at frequently replacing the warmed air layer at the tube with cooler air, then replacing it with more cool air, etc. Conducting heat from the tube and giving increased area also helps to move more heat to the air. Similar things are done inside the tubes on larger HX too. So we are trying to greatly increase flow tube turnover, but it is not really increasing turbulence in the engineering sense, as we are raising Re.

Anyway, anything that can cause the flow to turn over helps. While this mixing takes some energy and deposits that energy as heat, the effect is pretty small compared to the heat transfer that is occurring, and thus has little adverse effect...