Effect of ducted fan

[Tosho Otaguro]

Hi, everyone,

I am a half-retired, part-time engineer helping develop a fixed-wing plane with ducted fans inside a wing.  My role is to make a proper aerodynamic model, analyze it and optimize its geometries.

I've been building my preliminary model without a fuselage using a thick NACA4420 airfoil.  Here are views of my model in different angles.  Inside the wing I have two ducted fans tilted in 80 degrees against the chord.  The ducts are modeled using StackGeom with negative volumes.  In the ducts I put disks, and rotate them at the same speeds, typically at 6000rpm.  I used disks in order to simulate propellers at a limited computing cost.  The other parameters in VSPAERO are left as default.

I could obtain aerodynamic coefficients at different angles of attack which look pretty reasonable.  With Cp slices I see effects of the disks at the spanwise positions of the ducts.  However, in the wake trace visualization I could not see any effect of suction/blow-down of the fans to the wing surface.  I used OpenVSP v.3.31.1 by the way.

Questions

(1) Is VSPAERO able to include effects of ducted fans inside a wing?

(2) If yes, what is the most reasonable configuration to have a proper result?

(3) If no, is there any alternative to analyze this kind of non-typical aircraft?

I'd be very happy if someone gives any suggestion to me.  I'll be back on next Monday.

Best wishes,
Toshio Otaguro at Tsukuba, Japan

[Rob McDonald]

At a minimum, you'll want to use the panel method in order for negative volumes to work and to cut the hole through the geometry.

Next, the actuator disk model may not work well in a duct.  I would do some tests of a conventional axis-aligned actuator disk / duct combination to see how I felt about the solutions -- compare with what you can.

Next, the actuator disk is not very well suited for use in a strong cross flow (i.e. forward flight in your situation).  I would do some tests where I try to match forward speed and CT/CP appropriately to see how it does.

If either / both of these things do not work out, you'll need to go to an unsteady solution with rotating blades.  This will be a bit more challenging to set up -- and will also be considerably more computationally expensive.

Rob

--
You received this message because you are subscribed to the Google Groups "OpenVSP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openvsp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/cb6dccde-93ad-471d-ba4a-10d38330d9b0n%40googlegroups.com.

[Daniel E.]

I would suggest utilizing the aerodynamic solver FlightStream for such an analysis. They have validation cases for ducted fans and interfaces well with Openvsp geometry.

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/CAEppYpG5qWrDL0exoPQKSiODPjnr8ATq9SYHx%2BPdi1OTRUYicw%40mail.gmail.com.

[Tosho Otaguro]

Thank you very much for your very quick reply.  It's surprising.  Do you sleep?

Well, I will try to modify the model step by step, first by applying panel method.

I attempted to rotate propellers in the past, but could not obtain reasonable results.  Maybe, I used wrong settings.

Thanks anyway, and if you could find any direction, please let me know.

Best wishes,

Toshi

2022年11月21日月曜日 10:08:17 UTC+9 Rob McDonald:

[Tosho Otaguro]

Thank you very much for your response.  It's very helpful.

I am not familiar with FlightStream, and I will see what it is and how to use it.

Best wishes,

Toshi

2022年11月21日月曜日 10:22:33 UTC+9 deenri...@gmail.com:

[vivek ahuja]

Hi Toshio. As Daniel mentioned, you can also try FlightStream for the ducted fan problem. Like VSPaero, FlightStream is a panel method solver. We recently added some enhancements to the software to allow for the modeling of ducted fans using OpenVSP geometries. As Rob mentioned, you will need to model the geometry as a solid, but at that point you can export to FlightStream and set up an unsteady solver, as seen here:

Looking at your OpenVSP model, it looks like you should be able to make this transition to the solver in FlightStream.

If you need access to the software, let me know and we can set you up for it.

Best wishes.

[Tosho Otaguro]

Panel method has given me another flow pattern.

Here is the bottom view of the wing at an angle of attack of 5 degrees, at Re=5x10^5.

I see rise of surface pressure clearly on the downstream of the ducts.

What do you think?

Best wishes,

Toshi

2022年11月21日月曜日 10:34:34 UTC+9 Tosho Otaguro:

[Tosho Otaguro]

Thank you for your advise.  It looks very fascinating.

One thing about my role, I am not going to design propellers and ducts but to optimize wing-duct configuration to obtain good L/D in cruising as well as good hovering in take-off and landing.

Another thing is that I need a quick and light solution.  I started with xflr5, and now learning OpenVSP.
Thank you anyway for your suggestion.

Toshi

2022年11月21日月曜日 11:00:32 UTC+9 vivek...@researchinflight.com:

[Rob McDonald]

FlightStream uses essentially the same approach as VSPAERO to implement the actuator disk model.  They both use an approach proposed by Conway.

Conway derived an analytical solution for an actuator disk in isolation (with axial flow).  He also used it on the RHS of the potential solution to effectively put the actuator disk into the freestream component of the flow.

This means that you essentially have one-way influence from the actuator disk flow to the main flow -- but nothing in the main flow actually effects the flow induced by the actuator disk.  

I can believe that this might work out OK for an axially aligned ducted fan -- perhaps the influence of the duct can be captured separately from the influence of the disk itself.

However, when a disk is pointing up and there is a significant freestream velocity to the side -- the slipstream should start vertical and curve aft gradually.  I.e. the disk flow is influenced by the freestream in a way that the simple superposition of Conway's solution does not allow.

Conway has some additional work extending his disk solution to flow at slight angles -- but this does not extend to an 80 or 90 degree disk.

To capture this situation in FlightStream or VSPAERO, you will need to model unsteady rotating blades.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/CAAJ%3DGTdOGVyAaAbn9qrw5vJ%3D7LQ3H34cwSQvEE9s4aGnuJCtpA%40mail.gmail.com.

[Rob McDonald]

I do sleep -- quite a bit in fact.

Fortunately, answering most questions on this list can be done pretty quickly for me -- I generally find it is easiest to send out an answer as soon as I see a question.

The tricky part for this problem is going to be the geometry -- specifically, getting the propeller to exist 'inside' the negative space of the duct.

OpenVSP does not have the ability to model arbitrary combinations of Boolean operations, just a simple model of positive and negative.

When OpenVSP combines the geometry, any negative volume cuts away any positive geometry.  Actuator disks represent an exception -- they don't have volume and could reasonably need to exist inside a negative space.

The best answer might be for me to modify the Boolean algorithms to make propellers another exception to the negative volume cutting...

There might be some other (somewhat complex) workarounds, but there are some esoteric tradeoffs.

All this is going to take some time and cooperation in order to work out.

FlightStream would suffer a similar problem because they can only work with the geometry they get from another program -- so you need to sort through all this before you can try any solver.

I suggest you build up some experience with blades and ducts (without negative volumes inside of them) and we can build from there.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/683a29ac-7674-4580-9288-1585e008c11fn%40googlegroups.com.

[Rob McDonald]

I suggest you place a Cp cutting plane such that it cuts through the middle of the duct.  Then, visualize the velocity vectors on that cutting plane to see if you think they are reasonable.  I suspect you will not see the expected influence of the actuator disk.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/9932b2da-4792-4864-8454-0482881b9f15n%40googlegroups.com.

[Rob McDonald]

I suggest you work with conventional propellers and ducts to gain some experience with the tools and leverage the greater availability of solutions to compare with.

The hard part about applying any potential flow solution to your problem is that these methods will not capture separation.  In cruise -- when the disk is not powered -- I would expect the hole in the wing to produce substantial flow separation (and drag) that will not be properly captured by a potential flow code.

Similarly, the duct will cause substantial distortion and non-uniformity of flow into the fan when in forward flight.  These details are going to be very difficult to capture -- and very difficult to design around.

Rob

To view this discussion on the web visit https://groups.google.com/d/msgid/openvsp/17809359-6e0e-45de-9a53-f9b9f70a5f69n%40googlegroups.com.

[Tosho Otaguro]

Rob,

Thanks for many good suggestions.  They are very helpful.  I am also relieved to know that you sleep well.

I have made Cp slices at a couple of spanwise sections, and it seems that the code did not properly capture the flow over the wing surface, which should be modified quite a bit by the suction and jet of the ducted fans.  Now I understand the positioning of OpenVSP and FlightStream.  I have the same concerns as yours about separation and a strong cross flow in potential flow analysis.

Fortunately, one of my colleagues has been working on a full 3D CFD of the flow around a wing with a ducted fan.  It took long for modeling, but she is now very close to a realistic flow pattern of a jet in a strong cross flow.  I will try to figure a way out to combine her results with my potential flow analysis, starting with getting used to propellers and ducts.

Thank you very much for your suggestions.

Toshi

2022年11月21日月曜日 13:44:09 UTC+9 Rob McDonald: