Flowvis questions

[Mizanur.Sheikh]

I came across this thread while I was aimlessly going through the web so I thought I would answer you guys.

First of all, I am 4th year masters student at the University of Southampton (in england) studying Aeronautics & Astronautics with a major in aerodynamics.

My masters dissertation is designing and testing a 3rd scale open wheel race car (technically a hill climb car as there are literally no regulations so we get to decide what ever we want providing it fits in the dimension box), and my group just finish flow vis testing our car yesterday.

Flow vis is UV reactive but I think that's due to the nature of the chemicals, it's not really used to be shown under a black light, doesn't really help that much but it's made from a mixture of paraffin and ceramic powder and possibly a few other things.

Onto reading flow vis, people who say it's useless are fooled, the only way to get a more accurate picture of what's going on with the flow is to do PIV testing which is expensive and requires optical access to be used i.e. if the camera can't see it it wont get any data from it (and only works in a 2d plane) and every single F1 team uses flowvis on there wind tunnel models and there actual cars to gauge what's going on.
But when it comes to reading flowvis, it's down to aero knowledge, the flowvis plots the streak lines and it can show which way the flow is going, a relative speed, separation, vortexes and other things too.


Here are a couple of pics from my tests

This is off the rear wing and the last element of the rear wing, it shows a clear region of separation on the trailing edge, now separation = drag but, if you look at the flow vis on the inside of the endplates

you can see that the streak line is still very close to the aerofoil meaning the separation bubble is very small.
The endplates also show that air in front of the upper plane is being sucked down and below the upper plane which mean there is a region of very low pressure on the aerofoil which is a very good thing but ye, i could write a fair bit about what those 2 pictures show and that's just 2 pictures of a whole car.

So to answer the question 'how to read flowvis' aero knowledge and experience utlimately

and here is a picture of the car we designed

[Mizanur.Sheikh]

and pitot-static-rake only gives you pressure, the pressure can't tell you if you have reversed flow or not

[Giblet]

Nice post, and welcome to the forum.

I hope to see some more as you do some more testing

[Mizanur.Sheikh]

We just finished!! finished yesterday and had to clean up the flowvis mess from the tunnel today! I'll upload a video to youtube, it's quite techincal

our car is a 3rd scale tested at 30m/s with a rolling road system and the readings are scaled up to a full scale car at 120mph and we got about 7500N of downforce and L/D of 4.5 ish, a friend of mine that works in maurissa racing says they get a L/D of 3.9 ish

and we only have 5 days till the final report is due in and 6 weeks till university finishes :S

how to clean flowvis off the rolling road of a wind tunnel

http://www.youtube.com/watch?v=p_f_GNrj ... e=youtu.be

and as for the time for the flow vis to dry, about 10 mins you want to be running after applying other wise you just end up with brush strokes

[Rideway]

From the FI vs Lotus vs Aerolab trial:

The initial Lotus model produced 223.3 points of downforce with an aerodynamic efficiency of 2.512. By comparison the Force India model to which Aerolab had access as at July 2009 (which did not represent Force India’s most advanced geometry for the reasons explained above) produced about 272 points of downforce with an efficiency of 2.87.

[hardingfv32]

(Flowvis) it's made from a mixture of paraffin and ceramic powder and possibly a few other things.

1) Is it a retail product or did you formulate it yourselves?

2) Exactly how is it applied? Brush or spray? How thick? Timing of application: how near to the time the flow/tunnel will be started?

Brian

[Mizanur.Sheikh]

1) erm, i'm not sure tbh, they had some mixed in the wind tunnel already.

2) we used brushes but ideally it would be sprayed on because you leave brush strokes with a brush which can than cause confusing flow patterns were is strokes and not actually flow. You have a few mins before applying till it starts drying, you want to get it done in a bout 5 -10 mins (with a pump spray it would make it alot easier to do that) then run the wind tunnel at speed straight away for a few mins till you stop seeing flow moving.

[ESPImperium]

From the FI vs Lotus vs Aerolab trial:

The initial Lotus model produced 223.3 points of downforce with an aerodynamic efficiency of 2.512. By comparison the Force India model to which Aerolab had access as at July 2009 (which did not represent Force India’s most advanced geometry for the reasons explained above) produced about 272 points of downforce with an efficiency of 2.87.

A bit OT here, feel free to move their mods:

Do you have the court papers or a link to them?

Wading thru the legal bull, i think it would be interesting to read them.

http://www.blm-law.com/2114/14370/objec ... racts.html

[Rideway]

Do you have the court papers or a link to them?

I got it from here:
http://www.racecar-engineering.com/arti ... t-in-full/

[gixxer_drew]

I still dont understand the huge value in flow viz that it is used as much as it seems to be in F1. Using it n a wind tunnel is totally different thing to me with all the controlled circumstances. Wind tunnel is really tough to visualize flow in smoke diffuses really quickly so CFD is the best tool to me. Flow viz is restricted to a surface only. I know they are working on CFD correlations but it seems really tough to get good data like this. The photos are really interesting though.

The biggest thing I see is that the way the flow will visualize is 1000% different at 100kmh compared to 200kmh and you have no way to know at what speed a shape was generated. In my early days studying aerodynamics we did a lot with patterns of dust deposits on the side of land speed cars. I learned a lot from that I will admit, but once I had nice CFD and really expensive wind tunnels at my disposal, it felt like using a crane to pick up a cup of coffee.

If I was to do flow viz with oils, etc I would need controlled circumstances and onboard cameras with accompanying DAQ. You would need drivers to hold a certain speed down a straight away and then offload the data and even then you gotta look at exactly where stuff was puddled up before you gave it some real speed.

I did "poor man" PIV a few years ago working on my solver processes which was using dust particles and cameras on land speed cars. It was really about getting correlation with how the eddies formed and what I really learned from it is that it changed all the time depending on the shape of the car in question but I learned a lot about how things like to form. Maybe that knowledge is the core thing? I integrated what I learned and I did improve my absolute accuracy, but what I really learned was to just developed my CFD processes for best relative accuracy. Absolute accuracy seemed totally illusive to me, I could be 8 years or more behind these guys though seemed easier to get that in the tunnel. I'm glad someone is trying though. Can't wait for the "auto optimize my car shape" button in Fluent 2037.

[volarchico]

...
The biggest thing I see is that the way the flow will visualize is 1000% different at 100kmh compared to 200kmh and you have no way to know at what speed a shape was generated.

To take from the playbook of "Brian" I'll ask a question: What makes you think the flow will be so different at different speeds?

From my experience, unless something drastic is happening with regards to tripping turbulence at a new location, the flow at 100 km/hr will be nearly identical visually to the flow at 200 km/hr. The Reynold's number will be double, but by 100 km/hr you're well into the turbulent region anyway.

[gixxer_drew]

...
The biggest thing I see is that the way the flow will visualize is 1000% different at 100kmh compared to 200kmh and you have no way to know at what speed a shape was generated.

To take from the playbook of "Brian" I'll ask a question: What makes you think the flow will be so different at different speeds?

From my experience, unless something drastic is happening with regards to tripping turbulence at a new location, the flow at 100 km/hr will be nearly identical visually to the flow at 200 km/hr. The Reynold's number will be double, but by 100 km/hr you're well into the turbulent region anyway.

Hmmm I'm not exactly sure where to start here, but if you accept the concept of reynolds number I am not really sure how you could argue that the flow separation would be the same at two very different reynolds numbers?

But we can put that argument aside. Lets presume your flow is identical, the pressure forces will not be the same. Oil is not massless, it will be effected differently by different velocity and pressure of the air. It takes forces to move the stuff around and those are not an on off switch between in motion and not in motion.

Here is a real world example. You are testing the bottom side of a wing lets say and one test is 50km/h. At this speed the air only drags the oil (or whatever) 30% of the way up the nearly vertical flap, then you slow down even more and it runs back down to the bottom straight away. Try the test again at 200km/h stable in the wind tunnel and it goes all the way to the seperation point. Unless you have a car following or a camera hanging off the back how do you determine if that looks like an eddy formation near the end plate is an eddy formation or just that bunch of oil ran back down in some low speed corner?

When the car is sitting in the paddock, the stuff doesnt move at all. When its going real fast it moves all over the place... so every speed in between the two is a shade of gray and its moving more or less between where you are trying to correlate to.

Don't get me wrong, you can pick apart any test method until it seems like the problems are as such to make it useless, what you do in the real world is know the problems and try to deal with them. What I am saying here is that unless you are carefully controlling your speeds when you do flow viz, it doesn't seem all that useful to me and so I would tend to do it all in the wind tunnel where I have direct control. Unless you are trying to do correlation stuff... that I save for proving grounds. That may not be allowed for F1 though, so just generally speaking. The tool of flow viz seems to have a lot of error, from what I know of their wind tunnels, the error is far less than what I think would be realistic to achieve with something as crude as speeds presumably controlled by a drivers foot over the course of a practice lap. I doubt he has .01% or even 1%. I cant even find wheel speed sensors to go on race cars as accurate as the air speed controls in the tunnels I have used and those are a whole order of magnitude less accurate than F1. So you need a pitot tube setup, and a feedback system in the car to control the speed, but pitot tubes are noisy in the real world too noise isolation is a big deal and then filtering and everything. Or you set it for what it is on one part of the track, then theres a breeze when he comes back around or the temp is 1 degree cooler on the opposite end of the track or something.

Imagine you spend day in and day out professionally, calibrating sensors in wind tunnels, throwing out results straight away with 0.1% problems. Running CFD simulations on super computers about how the air is flowing through every little nook and cranny of your tunnel then checking them again by hand trying to get accuracy on something to multiple decimal places on something as simple as the speed of the tunnel or error in the belt vs air speed because you know it will make you throw out a month of tests... then you see someone doing a test with no speed controls to speak of, you just scratch your head and say "what are they doing here?". It's something I dont understand, that is for sure.

[Paul]

But how often do you see parts brought to the track and abandoned after they fail to show same promised as they did during CFD and wind-tunnel testing? In my opinion, as long as cars have to compete on real tracks, there will be no substitution to verifying results on the same track.

Sure, there is noise problem, but it might actually help find a problem with cross-wind or exhaust flow sensitivity that wasn't discovered during simulations. I imagine if teams run cars at normal speeds, they will pay attention to bigger differences between what flow-vis shows and what their simulations show; and when they run cars at fixed speeds, they will look into more minor details...

[gixxer_drew]

But how often do you see parts brought to the track and abandoned after they fail to show same promised as they did during CFD and wind-tunnel testing? In my opinion, as long as cars have to compete on real tracks, there will be no substitution to verifying results on the same track.

Sure, there is noise problem, but it might actually help find a problem with cross-wind or exhaust flow sensitivity that wasn't discovered during simulations. I imagine if teams run cars at normal speeds, they will pay attention to bigger differences between what flow-vis shows and what their simulations show; and when they run cars at fixed speeds, they will look into more minor details...

Well see, you dont want noise, you want precision and you want to be measuring real things like a cross wind and not the limitation of the equipment... you want to be studying that data to understand what you didnt discover in your simulations. If the sensor noise is larger than the accuracy of everything else, its useless. Your test lacks the precision to be useful.

If I am measuring thermal expansion of a 10cm cube of aluminum and the smallest increment I can measure is 1 cm I will conclude it does not expand. If I can measure accurately down to 1x10^-20 mm I can come up with a really in depth explanation for the result of every factor in play and as I study more with that same tool I will learn more and more about all the external factors, but if I dont have that accuracy I am lost. If my measuring tape expanding at a higher rate than the aluminum and I dont know it, I am going to conclude the same block is shrinking, opposite of what is really happening.

The whole point is to improve your car..... If you run 4 seemingly identical flow viz tests two of them correlate perfect to your wind tunnel and CFD and 2 of them are way off then what do you do next? Change your wind tunnel to match the two tests that didn't correlate? You can end up wasting your precious development time and budget and that must truly be the heart of this game. You want to understand all the complex interactions on a never ending quest for ever finer detail of your understanding. People have learned how to compensate cross winds and how to measure them happening to compensate, we are down to way smaller details than that at this point surely.

So what they are trying to measure here using flow viz, surely cannot be something precise, the test lacks the precision. So what is it? I am still wondering that. Its some kind of a general trending thing, so I just wonder what is going on. I can only guess.

[volarchico]

So do all the F1 teams just have a thing for oil painting? I think Flow vis has a place in testing, otherwise they wouldn't waste their time with it. Sometimes living in a lab or wind tunnel environment where the search for ever increasing accuracy skews a person's perspective until they forget that reality is out there and reality is not clean, accurate, or repeatable. Not every test needs to be precise to the 4th decimal to give useful information or to validate concepts.

How often is this seen when a car loses a bit of front wing endplate or a strake from the diffuser during the race and continues to lap at the same pace? Losing that piece should adversely affect the aero, but it doesn't within the error bounds that are the reality of a race.