CFD for Thesis project

Gildasd

Space fantasist
Joined
2 January 2013
Messages
69
Reaction score
2
Website
scifimeanderings.blogspot.com
For my thesis project, I am looking for a free (I'm a student and POOR) CFD program.
Right now I'm using http://flowsquare.com/ for 2D approximations and visualisations.
This program has to be for compressible fluids. Our University has advanced software, but only for incompressibles (water) to test hulls and props. Any add on's price can be calculated in Corvettes Z06.
There literally 100's of CFD programs out there, most seem tuned to one need. Sorting this out for the uninitiated is a bit hellish.

I only need to model a 1x1x1m cube and some turbulence generators (another cube to simulate a crane, container - etc)
The program must, above all, give me numbers, ballpark, for drag (or force) - that's all - at different angles.
This is just to establish my premise for my thesis and illustrate the first pages.
I can't use hacked software for this...
3D is a plus.
Images is a plus.

I have Win7 64bit PC with 16GB of RAM and cannot afford anything else. All my funds will have to go in the test models (8 sensors, 2 Arduinos, WIFI shield, etc per cube) for a 3 month outdoor test.
Then data collection on a vessel in the North Sea (fingers crossed).

Thanks!
 

Attachments

  • figure-000003500.jpg
    figure-000003500.jpg
    494.8 KB · Views: 143
Computational Fluid Dynamics

http://en.wikipedia.org/wiki/Computational_fluid_dynamics

Wind tunnel in a CPU if you will. Wifey would be terrible put off if I build a real one in the living room.

cfd2.jpg
 
One presumes from your posts you are based in America? If so, what about contacting one of the Indycar teams? They'll use some CFD but not as much as F1 so maybe you can beg a little time on one of their machines? Best of luck!
 
shedofdread said:
One presumes from your posts you are based in America? If so, what about contacting one of the Indycar teams? They'll use some CFD but not as much as F1 so maybe you can beg a little time on one of their machines? Best of luck!
Not to disparage your sleuthing, but I am French living in Belgium.
If I was American, there exist a couple of Academia options that were created by NASA.
However, being a "forrin" I cannot request machine time nor code. Weapons proliferation and this and that you see.
In any case, even if I did, my fluid speeds are in the 5 to 15m/s range, so a bit below the range of tools created to test things that stall at 29m/s - and might go supersonic.
As for racing teams, they have other things to do and cannot rerun results at the whim of my Thesis Master.
 
DSE said:
While I can't supply a source, I think you may need to think through your requirements a bit more thoroughly, or at least how you present them.
- Why do you need a compressible flow solver? What is fluid, what is the velocity?
- To what level of accuracy and uncertainty do you need to determine the drag coefficient?
- Is the flow laminar/ turbulent and/or do you need to be able to predict transition?
To validate my premise, get some numbers that I can check in real life. The fluid is air at 15°C, the velocity is 5 to 15m/s.
The accuracy level needed is quite low, as these tools go, tens of grams at most. It is beyond me to build a 360° test jig meant to reside outside, with anything more than that precision.
The range would be 0.01kg to 50kg of force (or the Nm equivalent). I know it will be a min and max with an average.
I wish to run the test in a flow that is initially laminar, that will become must turbulent upon hitting the nice right angles that will be provided.
Where and when transition occurs is of no importance, only the total drag figures.
 
As someone else asked, why do you need a compressible code? You are working with effectively incompressible air at such a low Mach number.
 
gwiz said:
As someone else asked, why do you need a compressible code? You are working with effectively incompressible air at such a low Mach number.
I took that most dangerous of liberties; "assuming"...
As air is a gas at the temperatures I am concerned with, I naturally assumed that I should work with a compressible gaz CFD program.
My knowledge in this is near null (I'm studying to be an Engineering Watch Officer - not to be an Engineer) so any advice is warmly greeted and taken to heart.
One thing, I've noticed that for a given "wind tunnel velocity", the air in accelerated to over twice that speed around the corners of the cube.
 
Hobbes said:
Swansea University is worth looking into. They supply CFD services to the Bloodhound SSC land speed record project.
http://www.bloodhoundssc.com/sponsors/swansea-university
Not much use even if they did it. I must have control, I can't have someone do it for me.
Practical reasons and thesis rules that I'd rather not twist.

I must have a program.

The Swansea Uni program, in it's full version, is classified. It's "light" version is worth millions.
It also needs a render farm, a bunch of boffins and some PhD students to run it.
Their special subject is the transonic region. Hardly ever reached by a vessel, unless you count the bits flying about after being hit by an Exocet.
A slight overkill for a 1x1x1m box at 5m/s...
 
bipa said:
If you want an open solver, you could try this:

http://code-saturne.org/cms/
Thanks, I'll check that.

On the 3D flow simulation, Blender, of all things, does it. But does not provide numbers. However it will be able to simulate the vortices at the corners.
http://www.blender.org/features/

In fact, aside from my peculiar quandary, one could set up a few easy tools to test how remotely viable some of the weirdness posted here is.
Would love to run 1914 to 1919 airfoils through the wringer.
 
DSE said:
By training and profession I am an experimental fluid/gas dynamicist. I've worked hand in hand with CFD specialists for over 25 years.
(1) I'm not completely sure you have properly described the problem to us and (2) I'm not sure you really understand what you are asking for.

Let's start with the problem set up first. To me it sounds like you want the drag on your cube in the wake of something else, ie your "crane." Sounds more like a wake flow problem than the effect of some small turbulence generator. If this is not the correct interpretation please correct me. What is the relative size of your turbulence generator?

Now, forgetting about the wake flow, what you ask for with all but fully laminar and attached flow around your cube is not a trivial problem to accurately compute in 3D. If one adds the need to have to prediction transition from laminar to turbulent flow then the problem difficulty just increased significantly. I may be wrong, but this sounds out of line with something some free push button code is capable of computing accurately. You may get answers, but their validity might be quite questionable.

If I have correctly interpreted your problem there may be simpler old school methods to come up with a prediction of the change in drag, say by coming up with a modified velocity for the wake flow to apply to a simple drag calculation using drag coefficients published in the literature, such as the Horner drag references.

The core of my thesis is:
Can we use simplified CFD modelling (or it's math equivalent) to predict air forces on a large irregular object in irregular winds: the superstructure of a ship.

Step one:
Model the most simple yet relevant object: a 1/1/1m cube.
Run the CFD on it at 5° intervals, each at wind settings from 1 to 15m/s (1).
With this data construct a "grid" where wind speed A at angle B should generate force C (2).
Run some tests with the clutter on the roof (couple of masts etc).

I fully intend to do the calculus methods. One of our maths professors is a bit of a wizzo.
I have a CAS calculator that I intend to test to the fullest on this. Maybe write a bit of code.

Step two
Build two cubes in foam each mounted on 4 load sensors.
Each cube is placed in a windy and exposed position that resembles the position of the superstructure on a ship (roof of university).
Both tests subjects has a wind velocity sensor, wind direction, barometric sensor and air humidity sensor.
All the data is stored locally and remotely during a 3 month period
With this data construct a "grid" where wind speed A at angle B should generate force C.

Step three:
Compare the lot. Gain a aneurysm.
If time is sufficient, test data on a large ship with a precise data logging system (MV Wandelaar maybe).

This might seem all very basic for aerospace people, but this kind of info is hard to come by on vessels.
We have some data on the aero of exposed hulls, but in this era of marginal gains, having data for the superstructure would be nice.
If I could create a simplified system to get a quick approximation, that would be wonderful...
This can be used to improve ship positioning systems, predict dead zones were gases can cumulate etc.
But some of this is already beyond the scope of my thesis.

I'm still making this up as a go along. My first tests have blown some of my original ideas of the water...
I'm running simulations to get more data, but each one lasts 8 hours...



(1) The forces are going to be minimal below 5m/s, but better run them anyhow. I could be wrong.
(2) This is a bit of a bore for experts, but I don't want to create a monster at first. If I can see that I can add extra data, it will be layers on top of something that works.
(3) Real test is a must, "real wind" is a fickle multidimensional wave form, not a neat flow.
 

Attachments

  • Principe bloc.jpg
    Principe bloc.jpg
    344.6 KB · Views: 72
Edit... please don't take this reply personally. Just struck a nerve and a personal pet peeve of mine. CFD is a good tool, when used appropriately and fully understanding its limitations. You most likely can get some answer to the mathematical as posed within the program. How well that problem represents reality is up for grabs.

No problem whatsoever. This is meant to be a learning experience for me too.
I had a notion of "CFD for all" due to having been in companies with Solidwork, where the tool was just a click away.
Now I've gone and seen what this costs - oh boy!
In my previous line of work, I've only used CFD to check convection current within museum cases and the like. Never an "active" flow.

In any case, learning how to do a "back of the envelope" approximations because software is not infallible:
One on the last jobs I did, a temporary viewing bay in an art museum, I used Solidwork to calculate the ground pressure:
On Friday it gave me a set of numbers. On Monday, I loaded the same exact file as saved on Friday and got a number 50% greater!
Doing a rough estimate would have spared me a lot of evil looks from the commercial people...

My background reasons for this thesis are:
- Learn more about aero and the maths behind it.
I don't want to be an aero guru, but If I can learn how to communicate my needs to him in a precise manner, then job done.
- Work with sensors and microcontroller (I might mix different flavours just to make it harder)
 
It's the biggest shape I'm allowed to plant on the roof...
A vertical rectangle, 1x1x2, would have been much better.
The only break I managed to wrangle was a 1m high table to put the cubes is the highest velocity flow.
(higher than the drawing)

The fluid is air. The CFD will be 15°C, or whatever is the historic average Temp and humidity over the test period.

The velocities are a spot of bother.
I've noticed on my simulations that at 5m/s the flow can accelerate to twice that.
But these are the wind speeds at which work boats (tenders, cable layers, dredgers...) operate.
 
For simple structures you can probably find tables / formulas that have been verified experimentally.

They still use wind tunnels for ships and buildings too, though the upkeep costs mean there are less and less of them around.
 
mz said:
For simple structures you can probably find tables / formulas that have been verified experimentally.

They still use wind tunnels for ships and buildings too, though the upkeep costs mean there are less and less of them around.
Yes, but the data we have in school is for cargoes, tankers, liquefied gas carriers. Mostly for ships over 150m long.
It's mostly geared to when a vessel is "light ship" (no cargo) and presents a massive wall to the wind.
The superstructure is nearly negligible.

My study is focused towards "work boats" that have a more or less constant freeboard (part of the hull that's out of the water).
These vessels are smaller (in 99% of cases) but are "all superstructure".

These vessels also use a lot of diesel, usually the most expensive kind, just to stay in a position (last boat I was on used about 350 litres per hour and it was small).

My thesis's end goal would be to compare my "cube" data to a real ship with a position keeping system that logs forces for relevance.

I'll be working on these kind of ships in the North Sea, so even if my work is not going to change the world, it might have an effect on mine.
Just the practical part, sensors, data logging, has value.

As for wind tunnels, we don't have combined wave + wind tunnels right now in Belgium. It's one or the other.
However, apparently one is being built in Oostende but it's too late for my thesis.

On a side note, the chitchat here has made writing my brief much easier, so thanks if only for that.
G.
 

Similar threads

Back
Top Bottom