On the Feasibility of Using Open Source Solvers for the Simulation of Air Flow in a Dairy Building
3 views
Formal Metadata
Title 
On the Feasibility of Using Open Source Solvers for the Simulation of Air Flow in a Dairy Building

Title of Series  
Author 

License 
CC Attribution 3.0 Germany:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. 
DOI  
Publisher 
WeierstraßInstitut für Angewandte Analysis und Stochastik (WIAS), LeibnizInstitut für Atmosphärenphysik (IAP)

Release Date 
2019

Language 
English

Content Metadata
Subject Area 
00:00
Uniform boundedness principle
Projective plane
Drop (liquid)
Mathematical model
Cartesian coordinate system
Resultant
Mach's principle
00:23
Metre
Goodness of fit
Hausdorff dimension
Projective plane
Normal (geometry)
Resultant
Open set
Mach's principle
01:47
Axiom of choice
Statistical hypothesis testing
Group action
Theory of relativity
Turbulenztheorie
Mass flow rate
Velocity
Turbulenztheorie
Mathematical model
Cartesian coordinate system
Measurement
Statistical hypothesis testing
03:44
Metre
Scale (map)
Scaling (geometry)
Element (mathematics)
Sheaf (mathematics)
Sampling (statistics)
Mathematical model
Mathematical model
Line (geometry)
Element (mathematics)
Statistical hypothesis testing
Mach's principle
Flow separation
DopplerEffekt
Wind tunnel
Velocity
Sheaf (mathematics)
Vertex (graph theory)
Turbulenztheorie
Boundary value problem
04:35
Scale (map)
Volume (thermodynamics)
Mathematical model
Mathematical model
Infinity
Bilinear form
Statistical hypothesis testing
Mortality rate
Cartesian product
Mach's principle
Stokes' theorem
Stress (mechanics)
Large eddy simulation
Computational fluid dynamics
Turbulenztheorie
Block (periodic table)
Boundary value problem
Film editing
Subtraction
05:09
Volume (thermodynamics)
Mathematical model
Insertion loss
Infinity
Focus (optics)
Power (physics)
Local Group
Mathematics
Iteration
Directed set
Turbulenztheorie
Equation
Subtraction
Linear map
Scale (map)
Focus (optics)
Spacetime
Volume (thermodynamics)
Mathematical model
Bilinear form
Measurement
Element (mathematics)
Open set
Cartesian product
Flow separation
Stokes' theorem
Mathematics
Numeral (linguistics)
Turbulenztheorie
Large eddy simulation
Equation
Hex map
Element (mathematics)
Computational fluid dynamics
Block (periodic table)
Boundary value problem
Resultant
Finite element method
06:41
Metre
Building
Scaling (geometry)
Multiplication sign
Gradient
Mathematical model
Average
Greatest element
Time domain
Measurement
Wind tunnel
Velocity
Average
Autoregressive conditional heteroskedasticity
Boundary value problem
Condition number
Friction
Randbedingung <Mathematik>
07:34
Area
Bilinear form
Resultant
Mach's principle
Time domain
08:07
Metre
Multiplication sign
Set (mathematics)
Bilinear form
Tetraeder
Open set
2 (number)
Time domain
Mach's principle
Population density
Hausdorff dimension
Computational fluid dynamics
Length
Resultant
08:56
Zirkulation <Strömungsmechanik>
Pairwise comparison
Mass flow rate
Zirkulation <Strömungsmechanik>
Mass flow rate
Time zone
Measurement
Bilinear form
2 (number)
Mach's principle
Plane (geometry)
Causality
Symmetry (physics)
Velocity
Average
Order (biology)
Buffer solution
Invariant (mathematics)
Field (mathematics)
Resultant
10:43
Zirkulation <Strömungsmechanik>
Line (geometry)
Image resolution
Multiplication sign
Algebraic structure
Ordinary differential equation
Power (physics)
Mach's principle
Goodness of fit
Prediction
Velocity
Energy level
Quantum
Field (mathematics)
Physical system
Standard deviation
Theory of relativity
Algebraic structure
Total S.A.
Mortality rate
Line (geometry)
Prediction
Bilinear form
Measurement
Vortex
Sample (statistics)
Velocity
Arithmetic progression
Resultant
13:49
Projective plane
Mach's principle
14:11
Geometry
Modal logic
Slide rule
Lattice (order)
Physical law
Mathematical model
Resultant
Cartesian product
Mach's principle
14:56
Axiom of choice
Modal logic
Building
Mass flow rate
Commutator
Building
Feasibility study
Prisoner's dilemma
Lattice (order)
Mathematical model
Resultant
Cartesian product
15:58
Point (geometry)
Presentation of a group
Projective plane
Division (mathematics)
Focus (optics)
Time domain
Arithmetic mean
Centralizer and normalizer
Wind tunnel
Velocity
Turbulenztheorie
Turbulenztheorie
Resultant
Condition number
18:01
Goodness of fit
Observational study
Observational study
Presentation of a group
Complex (psychology)
Mathematical model
Resultant
00:00
having you here today and today I want to present some results from our now nearly 3 years of
00:08
ongoing joint project of 3 like it's institutes from droplets pub
00:15
was from the US and I'm from the ATB and it's a bit more about application of modeling and here's
00:25
some hardline 1st I wanna talk about the motivation why we do what we do or what we did how we did this what kind of results began but we thought also lot using this opportunity to talk about how the media and good project but how the work with 3 different institutes went so far what could be done better what we learned and this in lessons learned and then some odd look what before the plan to do so it's all
00:57
about the me coming from the ATB were dealing with problems with and related its emissions and mostly it's dairy cows which I headed in dairy bonds like this they are very large you can see here the dimensions nearly 100 meters long and 11 meters high at 35 meters wide and our goal our aim at AT B is always to measure the emissions that come out of these very very large norm bonds and that is now it's demands of problem because nobody can really accurately measure these emissions because we faced large openings so naturally you would say OK let's just simulate because all these
01:50
effects which you see that these are the problems we are facing because the sponsor directly coupled to the turbulent ambient weather conditions and so I really doing this with measurements as experience it's not straightforward so a modeling is a choice of which we
02:10
wanted to do the motivation of the work is to simulate the air flow pattern because the for the wellbeing of the animals inside the barn there it's important to know hold hold the velocities all the animals solely lost his inside the barn and the flow pattern in the bond is we have formulated 2nd is the same as the relation of the air exchange this is pretty important for for the farmers to know the exactly exchange sales of for the wellbeing of the animals and then have for future works you to simulate the transport of particles and also guesses like ammonia like and so on and this leads to this we need to transient simulations and usually in our community she do simulations on full initially ventilated Bonci have and the commercial source like this like C plus plus and so on and this sometimes can get very expensive if you do a gendered simulations because you need this highly paralyzed and sometimes you have to buy licenses for every year of the CPU which you use so we said let's do some tests with so so to see whether these are feasible for for our applications and to do so we did
03:47
some experiments in all atmospheric boundary layer windtunnel which we haven't ATB we made a 1 to 100 scale down model and we measured it and you can see
03:57
here the atmospheric boundary layer we have the info section of something like 10 meters was with roughness elements to really develop a turbulent inflow and we have the really precise and accurately the laser Doppler anemometer where we can measure the velocities which we dates for the bond which you can see here in 2 D we measured it and several vertical sample lines in the bond for the velocity and we wanted to use these experiments as a database for these simulations which we defend
04:36
and we use that the difference followers of all so to inhouse chords and 1 widely used evokes 1st was the US compressible atmosphere model with could teasing cuts approach troubles made by all sides 2nd there was developed is developed at the VS it's the Parliament's over and the 3rd which we use it to be is the for form which is I come to this
05:08
later 1st there are some hints and all skin
05:11
atmosphere it wouldn't you said before it therefore the compressible and incompressible NavierStokes equations suitable for different Floridians' written in Fortran an interesting what this is the a Council approach which makes it very easy to implement obstacles and for the simulations we use the turbulence model a large any simulations and for the subgrid scare we use the Smolenski model 2nd loss is
05:41
all was the palm on its own power mathematics and object oriented numerics and to the US by a focal you FIL focus on floor and transport convection dominated problems written in C plus plus several available options for find spaces p 2 p 1 civilized finite elements and I have for that meshing I think it was mesh which will was also very easy to implement them the thought which we used old form it's also see finally volume C. plus plus stimulants model we use any as with a 1 equation it is common to model and for the measuring all form capable of several measures structured and unstructured of for these results which irrational we use something like a snippy hex Ms. Wade can just important STL file and its automatically generated was also very country to do so this is where the competition the
06:44
main in the end we ended up with the scale of the wind tunnel 3 meters long 1 meters heights and the height of the building so what this 1 200 scared and it's 11 centimeters high at the roof conditions that
07:02
prescribed and let me just took the velocities we measure in the winter so the mean velocities we had 0 gradient or do nothing conditions the old money at the end and the art let's noslip all free to model on the bottom depending on the court and slip condition on the top the time interval was 0 2 0 7 seconds and after that we did it statistics and compute the temple average for the velocities and we 1st started in to indeed you can see here some
07:38
results for for the open form we see some tervalent behavior but it turned
07:44
out that the we were somehow although predicting the the eddies in this area which led to results which were not really fitting with the experimental results which you see here so this result was most likely to the 2 dimensional simulations as so he said
08:07
OK we need to simulated in 3 dimensions and we ended up with a setting like this I show you results for Parliament and also for all forms for our common we had something around 1 million so it's density of whether they can here 0 0 0 0 2 5 seconds simulated times as it before 7 seconds at all as the length inside 80 millimeters unstructured grids and read it on several CPU's
08:41
these are the settings for open form is consumer machine it's a bit finer in the region of the bonds 1 comma decimal 2 5 millimeters the sexy he tourists and used 100 20 CPU and yeah we can see here some
08:58
results for all performed I show this because you can see after 1 2nd the floor slowly turns into a fully turbulent this is why we skip the 1st 2nd order resides in only took into account the least 6 seconds which now the the Eisel services to see a 2nd invariance of the velocity of Q criterion just to visualize and I would be see the in in the causes the velocity so we did this 6 seconds for a for the perform also for the palm 1 and therefore the some it's 1st assimilating I hope we can present the results in the next year when finished and
09:49
Antonelli could do some quantitatively and also quantitative she comparing the we see here for the experiments again what we see here is the velocity vectors and here for open form this is the average velocity field in 1 plane in the symmetry plane and this just for for a 1st measure the comparison and what we see here the recirculation sewn inside it's a well worth simulated then we are free circulation so outside of which are also to be seen at every recording was an experiment and also more or less the the flow pattern of the main floor inside the buffers you have the acceleration and the the deceleration and a bit upwards floor and then again the acceleration of the Opel's quantitatively speaking and therefore the quantitative we compare the
10:45
sampling lines which we measure to for the experiments and compare these with the average velocities of each simulation code and as you can see here for the experiments for formant pummeled we always will be as close as possible as said to the experiments and a more or less we can do this what's made us all made me very happy for long form we have very very come close to this and yet which is this simply in line with the flow and as the bomb and this was 1 of our main goal is to simulated the total exchange rates of the bond since this is a important to us and we we can assume that free just rely on this information we can accurately assimilate the exchange of the whole system the coming to the MIT Alliance we made some progress for this this was always or problem because when we did this in 2 D we had totally different lines therefore the simulation was something like this totally different to these so this is a progress for us since it's still not perfect but for us we can live with this and again at the at the ordered off the bond we have better results here again or computing the exchange rates this is sufficiently good for us the and what's also nice to see is that there are some deviations for power would which we do not see for altered form but this is in the region in the industry circulation so on and I think this might be due to the great resolution in a PALM on because this is the this was way call by off the roof you so there might be some improvements if we refine the grateful for the problem there so
12:53
some summary resides all good agreement for velocity fields instead the bond power woman form give the same level of security added to 2D relations delivered live poor results due to the over prediction of what X structures so for 3 D simulations are absolutely necessary but then also very costly because of the measures they would really of like I think the biggest simulations I did was was 14 millions for the fines great concerning the performance because this was also to see whether some calls would be better or less concerning the speed their own form was around 4 times faster than the Parliament for us we you will see the results then maybe for the summer with the yeah and I also wanted
13:53
to say something about the history because this was of project was which was actually initiated in 2017 was there I had a
14:03
policy and was really that resides in I did know what to do I just came here are some questions could anybody help me and seems like yes there was a lot for those that
14:14
maybe we just try was a problem and and I gave him some also fights for the great law for the geometry and he really some pretty fast results I was impressed so we made some for the meetings and there was another guy coming from the US and then also or slide India a colleague of mine were also joining and needed to the plan to do this with all 3 calls us come on and all forms and that we presented all iPhones presented the latest results in 2018 and the 3rd in a
14:48
mysterious we were convinced was the results we thought OK really cool we do this and then focal was also then said OK if
14:57
we put so much afforded let's do a paper on the feasibility of using so solvers for the transient simulation of airflow Endara buildings and then we had a further look at the results so more guys from the and even while those from this came and he has so in the end something like 9 people were working on this to prison right now and also the really hopefully find the
15:25
outcome will be the paper which we plan to submit in April we do it in a in a paper which is for all Community the 1st choice commuters animatronics in culture f for our community this is really enormity because you I've never seen any publication there was trends in so as an so so as so this was really on due to this network and this the framework was really very helpful to do so otherwise I would never have met these people to do this lessons
15:59
learned to this the moral are soft lessons I think in a way of you started as a just for fun project so we didn't have really some defined aims to be an old pro judgment project management in the classical way no clear division of course all we did something like this something like this I think it could have been a bit more straightforward but started as just for fun I said bigger challenge as I see it was the communications because the you possibly balloon are not very far away but still the only community and communicated over email and we use yeah we met sometimes also the mean we met and lights age we met in the US but set of a central location for information might be a good option if you if we ever start something like this again because many many information was lost due to the Mets which should not everybody red maybe not everybody was in sending all was the recipient that needs to be keep off because B starter was some experimental values I did some further experiments they changed and not everybody knew this so keep it uptodate really good and always inform the others also very good more information is better than less information there's something like the point before end at last divinity think everybody of you knows this so future like we if I
17:30
I think we will focus on the inlet conditions because as I've shown we have an atmospheric boundary layer windtunnel which means we have really a turbulent that which we did not simulate and I think we could enhance the results very much if we do as some turbulence on the internet and these are some preliminary results which we want to put on the the computational domain which I've shown you before and I hope we can presenters' realities next utility in an estate yeah conclusions of stolen this from
18:04
last year's of fondles presentation but I think this election benchmark problems are important benchmark might be more complex than expected speeds thought in the beginning this was really the easier Standard experiment which we could apply in the simulations and have good results turned out was more complicated than we thought this ad results are always goods comes anything against this and there are always ongoing benchmark study is made to be continuously updated and as I said before is never provide a necessary framework for this collaboration and we are have to share more results with other institutes also for
18:46
the others to get around