Modelling turbulent fluid-, thermo-, and droplet dynamics in the Leipzig Aerosol Cloud Interaction Simulator (LACIS-T)
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Number of Parts | 20 | |
| Author | ||
| Contributors | ||
| License | CC Attribution - NonCommercial - ShareAlike 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 and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this | |
| Identifiers | 10.5446/35354 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language | ||
| Production Place | Leipzig |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
Leibniz MMS Days 201812 / 20
00:00
Drop (liquid)TheoryQuantum chromodynamicsDoppler-EffektMass flow ratePresentation of a groupGroup actionComputer animationMeeting/Interview
00:22
Point cloudDrop (liquid)Quantum chromodynamicsTurbulenceFluidMathematicsStandard ModelThermodynamicsTerm (mathematics)Directed graphEvaporationFree groupFreezingScale (map)Range (statistics)Wind tunnelMultiplicationToeplitz-MatrixPoint (geometry)Scaling (geometry)Mass flow rateRange (statistics)Presentation of a groupTerm (mathematics)Quantum chromodynamicsEntrainment (chronobiology)10 (number)PredictabilityGrothendieck topologyCondensationSkewnessDifferent (Kate Ryan album)ThermodynamicsMetreFluidPoint cloudGroup actionDrop (liquid)Computer animation
02:11
Directed graphCondensationAtomic nucleusVaporPoint cloudRange (statistics)TurbulenceThermal fluctuationsDrop (liquid)Content (media)Standard ModelCondition numberInjektivitätVelocityThermodynamicsCondition numberPoint cloudWater vaporDew pointAtomic nucleusCategory of beingMass flow rateLoop (music)Filter <Stochastik>Content (media)Independence (probability theory)Point (geometry)Sheaf (mathematics)Range (statistics)Thermische ZustandsgleichungTurbulenceDirection (geometry)State of matterDrop (liquid)DiameterCondensationWind tunnelThermal fluctuationsGastropod shellEntropyThermodynamicsMetreClosed setMoment (mathematics)Letterpress printingMeeting/InterviewComputer animation
05:10
Standard ModelPoint cloudFluidPolygonComputer animation
05:34
Randbedingung <Mathematik>Time domainThermodynamicsFluidStandard ModelCorrelation and dependenceKörper <Algebra>RootSquare numberCondition numberThermal fluctuationsVelocityPoint (geometry)Line (geometry)Cartesian coordinate systemMathematical analysisInfinityVolumeDivergenceNichtlineares GleichungssystemStokes' theoremApproximationScalar fieldQuantum chromodynamicsStatisticsPressureOperator (mathematics)Link (knot theory)Wind tunnelGradientCondition numberDirection (geometry)Perturbation theoryRootSquare numberMass flow rateTime domainPoint (geometry)Arithmetic meanStandard ModelMathematical analysisQuantum chromodynamicsScalar fieldVelocityBoundary value problemDifferent (Kate Ryan album)ApproximationThermal fluctuationsPerspective (visual)ModulformMultiplication signDifferential equationSet theory2 (number)FluidDrop (liquid)Sheaf (mathematics)Cartesian coordinate systemKörper <Algebra>Nichtlineares GleichungssystemWind tunnelPhysical systemLine (geometry)PressureMetreStatisticsCoordinate systemNumerical analysisPresentation of a groupLarge eddy simulationTwin primeFrequencyEvent horizonComputer animation
09:08
Standard ModelVelocityFluidThermodynamicsArithmetic meanPlane (geometry)Square numberVelocityThermal fluctuationsRootRight angleState of matterMultiplication signResultantPairwise comparisonDeterminantStress (mechanics)Different (Kate Ryan album)MereologyComputer animation
09:53
FluidThermodynamicsStandard ModelTurbulenceAlgebraic structureSurfacePairwise comparisonConnected spaceLine (geometry)Term (mathematics)PressureMathematicsDiameterDistribution (mathematics)Randbedingung <Mathematik>Drop (liquid)Quantum chromodynamicsSymplectic manifoldMassSphereVolumeCalculationTrajectoryCharge carrierKörper <Algebra>InjektivitätVelocityHeat transferGradientVaporMaxima and minimaLagrange-MethodeLoop (music)Mass diffusivityLiquidSurface energySolid geometryContent (media)Variable (mathematics)Population densityDivisorEquivalence relationTrailMassCategory of beingVelocityCharge carrierFluidDiameterArithmetic meanNichtlineares GleichungssystemNumerical analysisConnectivity (graph theory)Point (geometry)Standard ModelDifferent (Kate Ryan album)MathematicsForcing (mathematics)Mass flow rateMultiplication signDegree (graph theory)SurfaceResultantSheaf (mathematics)RotationGradientAlgebraic structureSquare numberSummierbarkeitStress (mechanics)InjektivitätWind tunnelWater vaporThermodynamicsThermal fluctuationsQuantum chromodynamicsStiff equationVaporDrop (liquid)Position operatorBoundary value problemInvariant (mathematics)TurbulenceOrdinary differential equationPairwise comparisonAdditionMaxima and minimaVolume (thermodynamics)EstimationLine (geometry)Distribution (mathematics)Körper <Algebra>SubgroupCondition numberMathematical analysisTime domainMereologyScalar fieldCalculation2 (number)Product (business)SphereMoment (mathematics)Time zoneMatching (graph theory)Right angleFunctional (mathematics)Nominal numberGroup actionEuklidischer RingMomentumZirkulation <Strömungsmechanik>Computer animation
17:13
Standard ModelWater vaporTheory of relativityGrothendieck topologyMetreComputer animationDiagram
17:58
Drop (liquid)Quantum chromodynamicsPairwise comparisonDistribution (mathematics)Standard ModelDiameterStatisticsCondition numberScale (map)Distribution (mathematics)DiameterMoment (mathematics)Boundary value problemStatisticsMultiplication signTheory of relativityWind tunnelScaling (geometry)FrequencyStandard ModelGreatest elementPosition operatorMortality rateCartesian coordinate systemSet theoryCondition numberPhysical lawLattice (order)Computer animation
19:09
Standard ModelRight angleMass flow rateComputer animation
19:49
Drop (liquid)Quantum chromodynamicsDistribution (mathematics)Pairwise comparisonStandard ModelStatisticsCondition numberScale (map)ThermodynamicsSymplectic manifoldModule (mathematics)Sheaf (mathematics)Correlation and dependenceTerm (mathematics)Flow separationDiameterStandard ModelMereologyTerm (mathematics)Cross-correlationAdditionModulformScaling (geometry)ThermodynamicsCondition numberSheaf (mathematics)Distribution (mathematics)Boundary value problemResultantOpen setMass flow rateLarge eddy simulationMultiplication signComputer animation
21:21
Associative propertyMeeting/InterviewComputer animation
Transcript: English(auto-generated)
00:01
Welcome to my talk. It's now a bit more applied CFD, not so much in theory, as this is about the simulation of the flow and the Doppler dynamics in the Leipzig Aerosolar Cloud Interaction Simulator. Many of you might have seen it yesterday. My name is Silvio Smafus.
00:20
Jens Vogtlander was the other guy who gave these nice presentations yesterday, and we worked together with Dennis Niedermeier and our head of group is Frank Strathman. Okay, this is the outline. At first, I give a short introduction to the topic, and I talk about the fluid and thermodynamics inside the wind tunnel.
00:41
The third point is about the particle and droplet dynamics, and finally, I give some conclusions and an outlook. But at first, for the introduction and about clouds. You all know clouds, I think, and you all may know that they play a huge role in long-term climate and short-term weather predictions,
01:01
so we should really investigate how they work. Unfortunately, this is hard because they are transient, they are turbulent, intermittent, inhomogeneous, and they cover a wide range of spatial and temporal scales as you can see here in this nice overview. We are in LASIS-T dealing with this site.
01:22
We have a look at the microphysics, how do cloud condensation activate to cloud droplets, how do they act after they activated to these droplets, and how the mixing of different flows contributes to this.
01:41
Of course, clouds have much larger scales. For example, entrainment in the scales of 100 meters, clouds can have scales of multiple tens of kilometers, and finally, you have covered almost all continents with clouds. Additionally, they are hard to reach
02:02
because they are up in the sky somewhere. You can use planes, but then they are not reproducible, for example, so as the clouds are hard to reach, we thought it might be a good idea to get them to us, and we set up this cloud laboratory to have reproducible conditions to investigate the cloud developments.
02:23
Clouds are usually formed by water vapor, which condenses a cloud condensation nuclei. These CCNs can be maybe, for example, dust, as we have seen yesterday in the talk by Robert Wagner. They can also be made of black carbon from industry and incomplete combustions.
02:41
Volcanoes emit sulfates, which act as CCNs. Even phytoplankton can act as cloud condensation nuclei, and feet deal with salt, which might be lifted up from seawater. These condensation nuclei have size ranges of approximately 100 nanometers, some less, some more,
03:02
and the condensation at these nuclei is mainly defined by the material properties, the particle properties like diameter and the ambient physical properties, and here especially the temperature and the water vapor content. LASIS-T is now designed to investigate how much do turbulent fluctuations
03:21
of these ambient properties influence the particle and to up with microphysics. Now some words about LASIS-T. It is short for turbulent light switch aerosol cloud interaction simulator. You see a schematic here. It's a closed loop grating entire wind tunnel.
03:42
The flow direction is clockwise here. It features two branches, each with up to 5,000 liters per minute of flow speed, air speed, and we can precisely define the thermodynamic and flow conditions in both of these branches independent from each other.
04:02
So the principle is that we have a twir here which supplies warm and dry air, which is then accelerated by two blurs, one for each branch. The air is led through filters to get particle-free air at first. The air can then be humidified to get a defined dew point, and afterwards this humidified air
04:22
is led through heat exchangers which are coupled with thermostats which set the humidified air to a defined temperature, and then here these two branches are mixed, and so we can reach, for example, super-saturated states here in our measurement section,
04:40
and additionally here in this point where these two branches mix, we can inject well-defined particles. These are these cloud condensation nucleus. At the moment, we use salt particles as already set, and these will form a cloud inside this measurement section. Now this talk is about the question
05:01
if we are able to simulate the flow, the thermodynamics, and the particle and droplet behavior inside this wind tunnel. Before we start, there's a not-so-good-to-see video of the cloud. This was taken from a TV show, Triset Nano. They made some nice videos of this. It's quite fast, but there's a slow-mo here.
05:23
You can see how these turbulent motions influence the cloud forming development. Now we come to fluid and thermodynamics, and first, some principle things to investigate
05:42
the influence of the turbulence on the microphysics. We need, of course, fields for instantaneous mean and mean square values, and measurements usually don't give enough information about all these quantities at once,
06:01
so usually you have point probes or you can measure along a line, but you don't get the whole field inside, so this is why we need the simulations, and we need exact knowledge of the thermodynamic conditions inside our wind tunnel before we can simulate the particles and the droplet activation, and this means velocity and turbulence,
06:20
temperature and humidity. A few words about the measurement techniques for the flow. This is the measurement section in detail. It's two meters high, 80 centimeters wide, and 20 centimeters in depth. The coordinate system has its origin here at the aerosol inlet, the y-axis being here in the short direction.
06:41
The z-axis is a long direction. This will be used throughout the presentation, and we use point probes for measurements, hot wire for velocity, and cold wire for temperature. They are usually then moved along lines in y direction, here 120, 130 centimeters below the aerosol inlet.
07:05
We do high-frequency measurements, approximately three kilohertz, so that we can get through an analysis to different quantities, root mean squares, and the means and the fluctuations.
07:20
Okay, for the numerical methods, we use open form. We solve the transient incompressible neighbor-stokes equations. For these examples, we apply a pimple algorithm which allows for larger time steps than PSO. The perspective solver is called pimple form.
07:40
For these examples, we used business approximation solvers for temperature simulation, and humidity is just treated as passive scalar. Use second-order schemes for almost everything, except these exceptions, and we consider turbulence with large ID simulations model, which is called dynamic K equation.
08:01
We calculated 2.5 seconds of real-time. But the statistics to compare with the measurements was only done for the time after 0.5 seconds to have a quasi-steady flow. The domain and the boundary conditions. Of course, we do not model all of the wind tunnels, so we don't need a twir and the blowers and so on,
08:22
so we only model this section here. This is an hexauter mesh with approximately 7.5 million cells. This is not the actual mesh. We all do some grid refinement at the walls. We have additional refinement here at the particle inlet
08:43
and so on. Here's the main boundary conditions. Here at the inlets, we can set fixed velocity, temperature and humidity. Under the outlet, we have fixed pressure. Here you can see a turbulence grid which introduces some artificial turbulence inside our flow,
09:03
and then at the front and back, we use cyclic boundary conditions. Okay, you see some first results from left to right for the instantaneous velocity, from mean velocity for the turbulent fluctuations and the root mean squares of the turbulent fluctuations, and this is done for two different planes.
09:21
This explains the different values here. One plane is directly below a bar of the turbulence grid. The other plane is in the middle between two bars of this turbulence grid. As you can see from the beginning, the velocity evolves, and finally, here these mean values reach some quasi-steady states.
09:43
You can see here the nice turbulent fluctuations we have in our channel and also these fluctuations reach a quasi-steady state after some time. Here's the comparison with measurements along these four lines I mentioned earlier in these four heads.
10:01
At the bottom axis, you have the Y coordinate, and the big Y axis, you have the velocity, and I think we hit the measurement results quite good. For example, we can really meet these dimples here in the middle, so I think we get the flow quite good,
10:22
and here you see some visualization of the turbulent structures we have in our tunnel. This is done with the Q-ISO surfaces, which is the second invariant of the gradient of the velocity, and from the beginning, you can see these structures evolving here downwards from the turbulence grid and then slowly decreasing downwards in our channel.
10:46
We compare this to our measurements with the turbulence intensity, which is the ratio of the turbulent fluctuation with mean square and the mean velocity. Actually, it's just for two components because also the hot wire measures
11:01
only two components of the velocity. You see the results for this here. Also here, I think we hit the results quite good. Even though we overestimated turbulence a little bit, but overall, I think this result looks quite okay.
11:23
Then for the temperature, this is an example where we have warm and cold air mixing. This is a snapshot of the simulation. You have 23 degrees on the left side, 13 degrees on the right side. You have walls with a certain temperature, and here you have a turbulent mixing zone.
11:42
This is now compared by means of mean temperature, which is made relative with maximum minimum value, so you have zero value for minimum temperature and maximum value as well. Here, you can see that we also hit these results quite good.
12:08
But again, for the turbulent properties, there's some overestimation here. Last thing for the fluid and thermodynamics is humidity. We treated humidity here as a passive scalar,
12:20
which is transported to the flow. Combined with the temperature, we get a relative humidity. You can see here, and this is now the interesting thing. You have here relative humidity of approximately 50 on the left side, and here approximately 70% on the right side.
12:42
But we have here some higher saturation, so up to 75% is where the turbulent mixing might play a role, which might lead to faster droplet activation, for example. And as I said earlier, this is what we want to investigate with our wind tunnel.
13:03
Now for the particle and droplet dynamics. At first, again, some words about the measurement technique. We use a white light particle spectrometer from Pallas. It's called VELAS, and we have two different measurement heads depending on the case. We want to investigate.
13:21
One is 10 centimeters, the other one 35 centimeters below the aerosol, and then we inject particles, let them follow the flow down there, and then they start to absorb water vapor. They start to grow or to shrink depending on the physical properties and the surroundings.
13:40
And then we measure the particle diameters here or here for 20 minutes, and from this we get the particle size distribution from our measurements. Okay, now for the numerical method, we use La Croix and Chen particle tracking, which tracks individual particles through the flow field. We have to assume that they are point mass particles,
14:01
so we need models for the interaction of the carrier fluid with other particles in electric fields, but here only the interactions with the carrier fluid is relevant. Usually, as usually, they are approximated as spheres, which makes things a lot easier. The particle size has to be, of course,
14:23
a lot smaller than the control volume size, and usually you need three ODEs to describe the particle behavior. The change in position is defined by the velocity. Change of velocity is defined by the sum of the forces acting on the particles, and you can also calculate the rotation of the particles,
14:41
but we neglected this here. This might not play well. The models we used, the forces acting on the particles here is just the track force. Everything else is neglected. Additional models, of course, we have an injection model defining the particle position, size,
15:00
diameter, velocity, and so on. We have a growth or mass transfer model, which defines how much water vapor the particles get from the gas phase, which is mainly driven from the gradient between the surface vapor saturation at the particle surface and the saturation
15:21
at the ambient air flow. Okay, this is some schematic of this whole thing. We have this Euler loop, LaGrange loop, and the growth model loop. Here's all the data we need for our growth model, so we have fluid data, particle data, model data,
15:41
which is all we need for the growth model. This might not be so relevant here. The only thing which is interesting is that we have the largest time steps for the Euler loop, then we have sub-loops for the LaGrange loop, and inside these sub-loops, we have again sub-loops for the growth model, as this is a quite stiff equation.
16:05
Okay, and the boundary conditions, I used the same domain as before, but now the only difference is in the solver, now the compressible Navier-Stokes equations are solved, everything else is kept the same. And we consider two cases here with high ambient region relative,
16:24
humidity as particles we use, sodium chloride as in the experiments. They have a diameter of 320 nanometers, which also matches the experiments. Of course, they are injected at a particle inlet.
16:41
They are tracked for approximately 0.5 seconds, which is just the time they need to travel through the measurement section I modeled, and the sub-time steps for LaGrange and tracking and the growth model are chosen automatically, so that the particle number is below 0.5, and the mass change is not more than 20 percent
17:03
of the particle mass for one growth model sub-loop. Okay, so we can come to the results for this. First case is this with a medium relative humidity.
17:20
At this side, you see this relative humidity, here the temperature. Here you can see the white values are comparatively low. Relative humidity from the aerosol inlet, so the particles are injected here with this 320 nanometers in diameter, which is bluish colors then, and they start to grow as the relative humidity increases,
17:42
so they get this reddish color here, and so we are able to reproduce this behavior that they start to grow and to absorb water if the relative humidity is high enough. So this is then evaluated here. It is positioned similar to our measurements,
18:03
and you can see here particle size distribution with diameter at the bottom and relative frequency here at the left side, at the left axis. The simulated diameter is not completely off, but it's a bit lower, and what you really can see very good
18:23
is that the simulations predict a really narrow distribution, while the measurements show that they are even ... that they have comparatively broad diameter distribution.
18:40
Why is this? We do not know at the moment. Maybe we do not really meet the boundary conditions we have in the wind tunnel. Maybe we also not have enough time for some meaningful statistics in our simulations. Maybe also the sub-grid scale turbulence might play a role, which was neglected for now, so this might also disperse the particles a bit more,
19:02
but we have not implemented a model for this yet. Okay, the second case with the high humidity. Again, on the left side, you have this humidity. Right side is temperature. Now the particle in that flow has a comparatively high humidity,
19:21
so that the particles start growing at the beginning, and then, again, they tend to shrink. As here you can see in this turbulent mixing zone, the humidity is mixing, so the air becomes twirier.
19:42
Well, this is also compared to measurements. Here the particle diameter is quite good, I think, but still the distribution is definitely too narrow. This might be for the same reasons as before,
20:02
but at least the diameter looks a bit better here. This brings me to summary and outlook. At first, for the summary, we simulated the flow and the thermodynamics inside LASIS-T with large eddy simulations in open form,
20:20
where the results agreed quite well with the measurements. We implemented the particle growth model in open form's LaGrangean model. The first simulations look quite good, I think, for a first shot, but of course there's some room for improvement. For the outlook, as I mentioned,
20:40
we want to implement a sub-grid scale turbulence dispersion for the particles. You want to do further simulations for other conditions for longer times and so on. You want to do more measurements, for example, also in the lower regions of the measurement section, which was not mentioned yet. And in the long term, of course,
21:01
we want to see some correlations between turbulence and the top-level particle behavior. Additionally, we also want to do some improvements at LASIS-T, for example, insulation and sealing to get the boundary conditions a bit better, and we also want to improve our particle measurement techniques. Okay, then. Thank you for listening, and I'm open for your questions.
21:27
Thank you very much for the nice presentation. Other questions?
Recommendations
Series of 2 media