All entries for February 2007
February 27, 2007
- MT03-01: TIR
It’s becoming clear that the central differencing is not making a huge difference. I will now try different turbulent intensities:
mt03-01-16: Ti=0.3 (will run these when the other sims have finished)
mt03-01-17: Ti=0.4 (will run these when the other sims have finished)
mt03-01-18: Ti=0.5 (will run these when the other sims have finished)
- RS01: Lille Bump
Now that we finally have a set-up for the bump we can start getting the final chapter underway. I also need to get the code estimating the wall shear stress, pressure distribution and other properties. I need to decide what main results I’m going to use.
I’d also like to have the literature review draft finished by the end of the week. I need to assess the major works in the area as well as put more detail in about the Lille work. I will put these into my list of goals to complete.
February 26, 2007
Starring George Clooney, this film went a long way (in my eyes) to repair the damage he did to his reputation with the farce, Batman & Robin. Clooney’s soulful representation of a psychiatrist whose harsh actions lead to his wife’s suicide is excellently played without too much drama and sentimentality being forced down the viewers throat. (As, sadly, is the case in many modern US films)
The director, Steven Soderbergh, produces a superb set with a tremendous attention to detail whilst the soundtrack by Cliff Martinez is truly inspired.
The story is an interesting one although it does leave a lot of questions open. When I first watched it I thought ‘What crap! I have no idea what happened!’ but the more I’ve watched it, the more it grows on me and I can start to see more behind the story. I’ve since bought the original book by Stanislaw Lem but have yet to get round to reading it (The other half has got me hooked on Terry Pratchett!)
All in all, watch this film; it will give you something to think about!
Martinez blends electronica with percusive vibraphones seamlessly and produces a very eerie but emotive sound. The chord progression throughout (4-5-6) is very persistent and seems to represent the constant pain that Kelvin feels at the loss of his wife.
The music has an excellent way of keeping the viewer on edge with the strangeness but there is also enough of the familiar to prevent it being too out of the ordinary.
In my opinion, one of the best soundtracks ever.
- MT03-01: Turbulent Inflow Rescaling
D’oh! I have set them all running with ncond=0 which means the default method is activated. This is a hybrid method but because the Reynolds number (and hence the Peclet number) is so high, this equates to upwinding. No wonder there was so much dissipation!
Will set running the following in response:
mt03-01-10: 05 but with ncond=3
mt03-01-11: 06 but with ncond=3
mt03-01-12: 07 but with ncond=3
mt03-01-13: 08 but with ncond=3
mt03-01-14: 09 but with ncond=3
Correction: ncond=0 is upwinding directly
February 21, 2007
Have currently got running a few sets of simulations…
rs01-02-A01: Lille bump laminar
rs01-02-A02: Lille bump RANS
rs01-02-A03: Lille bump 2D DES
mt01-03-04: 3D boundary layer using LES formulation & grid 3 (Ey=1.05)
mt01-03-05: 3D boundary layer using laminar solver & grid 3 (Ey=1.05)
mt01-03-06: 3D boundary layer using laminar solver & grid 4 (Ey=1.0375)
mt01-03-07: 3D boundary layer using laminar solver & grid 5 (Ey=1.025)
mt01-03-08: 3D boundary layer using laminar solver & grid 6 (Ey=1.0125)
mt01-03-09: 3D boundary layer using laminar solver & grid 7 (Ey=1.00)
These are all taking a while. Until then, I will get the plans sorted for the rest of the CFD module then get the marking complete for the posters.
Tomorrow, the plan is to print out Chapter 2, read the paper Yongmann gave me, copy Peter’s papers and download the other results.
February 20, 2007
For the record, here are the bike reviews in recent issues of 220 Triathlon magazine:
Issue 205 (March 2007): Scott Plasma Contessa vs. Litespeed Bella
Issue 204 (February 2007): Planet X Stealth vs. Orbea Tri 3
Issue 203 (January 2007): Felt DA vs. Giant Trinity C Zero
Issue 202 (December 2006): Focus Cayo Expert vs. Specialized Tarmac Comp
Issue 201 (November 2006): Cannondale Ironman Slice Six13 Si vs. Litespeed Saber
Issue 200 (October 2006): Triandrun Mirage vs. Koga-Miyata GranWinner
Issue 199 (September 2006): Trek Equinox 7 vs. Cannondale Feminine Synapse
Issue 198 (August 2006): Argon 18 Mercury Tri vs. Look 496 Tri
Issue 197 (Summer 2006): Carrera Vanquish vs. Dawes Comp Giro 500
Issue 196 (July 2006): Trigon EX1 vs. Litespeed Pavia
Issue 195 (June 2006): Fuji Team Pro vs. Merida Road HFS
Issue 194 (May 2006): Ribble Dedacciai vs. Focus Variado Expert
Issue 193 (April 2006): Isaac Efficiency vs. Scott Plasma Pro
Issue 192 (March 2006): Specialized Transition Comp vs. Trek Equinox 9
Issue 191 (February 2006): Giant OCR2 vs. Bianchi Via Nirone 7 Xenon 9v
Issue 190 (January 2006): Quintana Roo Lucero vs. Cannondale Ironman 1
Issue 189 (December 2005): Cinelli Xperience vs. Decathlon Comp 9.1
Issue 188 (November 2005): Kuota Kalibur vs. Felt S22
Issue 187 (October 2005): Cervelo P3 vs. Guru Crono
Issue 186 (September 2005): Condor Squadra vs. Orbea Vitesse
- MT03-01: Turbulent Inflow Rescaling
Have checked out the results so far for mt03-01-04 and -05 and they are more or less identical. This shows that the unsteadiness is not being killed by the eddy viscosity but more likely by numerical issues within the grid. To test this, I have set running the following:
mt03-01-05: Laminar using grid m3103g (Ey=1.05)
mt03-01-06: Laminar using grid m3104g (Ey=1.0375)
mt03-01-07: Laminar using grid m3105g (Ey=1.025)
mt03-01-08: Laminar using grid m3106g (Ey=1.0125)
mt03-01-09: Laminar using grid m3107g (Ey=1.0)
(Although -09 was killed due to lack of memory on Solar. Each sim takes around 1.8 GB of RAM)
February 19, 2007
Have sorted files into the correct places so that I have more room again now. Have deleted most of the development runs as I only really need the key runs and the final code to work.
Am currently sorting through the simulations and dividing them into development runs (Stages 1 & 2) and useful runs (Stages 3 & 4).
- MT01-01: S-A Model testing
Will take one of the wall functions sims which has a fine near wall grid and use this as mt01-01-A01. I will then try a number of different variables and record these simulations for interest. Will set this up now. (Prob use mt01-03-25 aka mt01-03-A01 first)
- MT01-02: DES testing
I will simplify these by concentrating on the Re=7500 case for now. Then once this is sorted, I can try the lower Re for completeness. Will do this at home later.
- MT01-03: Wall Functions testing
Have a good 12 runs for this (6 grids with & without) and just need to produce some results. Will do this at home later.
- MT01-03: Wall Functions Validation
Here are the latest results for the wall functions validation:
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a01to06uprof.jpg?maxWidth=500" alt='Wall Functions Validation- Velocity Profile without wall functions' title='Wall Functions Validation- Velocity Profile without wall functions' />
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a01to06unearwall.jpg?maxWidth=500" alt='Wall Functions Validation- Near Wall' title='Wall Functions Validation- Near Wall' />
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a01to06vissa.jpg?maxWidth=500" alt='Wall Functions Validation- Vissa profile without wall functions' title='Wall Functions Validation- Vissa profile without wall functions' />
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a07to12uprof.jpg?maxWidth=500" alt='Wall Functions Validation- Velocity Profile with wall functions' title='Wall Functions Validation- Velocity Profile with wall functions' />
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a07to12unearwall.jpg?maxWidth=500" alt='Wall Functions Validation- Near-wall Velocity Profile with wall functions' title='Wall Functions Validation- Near-wall Velocity Profile with wall functions' />
<img style="padding-right:5px;padding-left:5px;" src="/images/apreece/2007/02/19/mt01-01-a07to12vissa.jpg?maxWidth=500" alt='Wall Functions Validation- Vissa Profile with wall functions' title='Wall Functions Validation- Vissa Profile with wall functions' />
Note that the wall functions simulation for A09 (triangles) has some simple error causing the erroneous results.
- MT01-03: Turbulent inflow rescaling
Have set running mt01-03-04 which uses an LES formulation (ymin=dx*0.65). Have also set running mt01-03-05 which is a simulation without turbulence treatment.
- RS01: Lille bump without F.C.
Have set running rs01-02-A02 which is the same as rs01-02-A01 but with RANS. I may need to limit fv2 to make stable if necessary.
- The Lille Bump Set-up
Here are the final values of the boundary layer for the Lille bump.
d (at 15m) using fully turb bl approx= 0.227 m
d (at 15m) using transitional bl approx = 0.224 m
We shall use the transitional bl approx (although there isn't much in it) This gives the other boundary layer properties as:
d* (at 15m) = 0.02796 m
theta (at 15m) = 0.021764 m
Re(theta) = 1.496 x 10**4
wss (at 15m) = 0.144246 Pa
We cannot model the full Re with the resources available so will model Re/10 (u= 1m/s)
We want to preserve the boundary layer thicknesses at the inlet so, in essence, we will need to reduce the run-up length. This is done by finding the value of x which corresponds to d=0.224m. This is given as:
x (development length) =11.3 m
This gives the same boundary layer thicknesses (by definition) but gives a reduced Re based on mom. thickness and wall shear stress:
Re(theta) = 1.496 x 10^3 (1496)
wss (at 15m) = 0.002419 Pa (2.419 mPa)
This makes one wall unit equal to:
y+ = 1 > y = 0.328 mm
We plan to use grid spacings of:
dx < 120 mm (nx=300)
dy = 10 mm (ny=100)
dz < 42 mm (nz=50)
Which equates to:
dx+ < 366
dy+ = 31
dz+ < 128
Although the dz+ is greater than 100, this is only for the very edge nodes so I don't think it will be a problem.
From a recent boundary layer simulation to produce the boundary layer with the desired thickness, we get the wall shear stress:
wss = 0.00237 Pa
This compares well with the above estimates (around 2% error)
February 15, 2007
An excellent film which captures well the work of Frank Miller from which it is inspired. Christian Bale provides well developed Bruce Wayne/Batman character allowing the viewer to easily identify with his situation.
The film score is also superb with some interestingly contrasting orchestral pieces. All in all, one of my favourite films.