Tag Archives: Reynolds number

About the influence of viscosity: The Reynolds number

Leave a reply

This blog post was written for Elin Darelius & team’s blog (link). Check it out if you aren’t already following it!

—

I read a blog post by Clemens Spensberger over at  a scisnack.com couple of years ago, where he talks about how ice can flow like ketchup. The argument that he makes is that ketchup on your hotdog behaves in many ways similarly to glaciers on for example Greenland: If there is a layer of a certain thickness, it will start sliding off — both the ketchup off your hotdog and the glaciers off Greenland. After most of it has dripped to your shirt or in the ocean, a little bit still remains on the hotdog or the mountain. And so on.

What he is talking about, basically, are effects of viscosity. Water, for example, would behave very differently than ketchup or ice, if you imagine it poured on your hotdog or raining down on Greenland. But also ketchup would behave very differently from ice, if it was put on Greenland in the same quantities as the existing glaciers on Greenland, instead of on a hotdog as a model version of Greenland in a relatively small quantity. And if you used real ice to model the behaviour of Greenland glaciers on a desk, then you would quickly find out that the ice just slides off on a layer of melt water and behaves nothing like you imagined (ask me how I know…).

This shows that it is important to think about what role viscosity plays when you set up a model. And not only when you are thinking about ice — also effects of surface tension in water become very important if your model is small enough, whereas they are negligible for large scale flows in the ocean.

The effects of viscosity can be estimated using the Reynolds number Re. Re compares the effects of the velocity u of the flow, a length scale of an obstacle L, and the viscosity v: Re = uL/v.

Reynolds numbers can be used to separate different flow regimes: laminar flows for very low Reynolds numbers, nice vortex streets for Re > 90, and then flows with a stagnant backwater for high Reynold numbers.

Dependency of a flow field on the Reynolds number. Shown is the top view of a flow field. You see red obstacles, stream lines in blue (so any particle released at any point of a blue line would follow that line exactly, and in the direction shown by the arrow heads)

Dependency of a flow field on the Reynolds number. Shown is the top view of a flow field. You see red obstacles and blue stream lines (so any particle released at any point of a blue line would follow that line exactly, and in the direction shown by the arrow heads)

 

I have thought long and hard about what I could give as a good example for what I am talking about. And then I remembered that I did an experiment on vortex streets on a plate a while back.

Vortices created on a plate

Vortices created on a plate

If you start watching the movie below at min 1:28 (although watching before won’t hurt, either) you see me pulling a paint brush across the plate at different speeds. The slow ones don’t create vortex streets, instead they show a more laminar behaviour (as they should, according to theory).

Vortex streets, like the one shown in the picture above, also exist in nature. However, scales are a lot larger there: See for example the picture below (Credit: Bob Cahalan, NASA GSFC, via Wikipedia)

Vortex street. Credit: Bob Cahalan, NASA GSFC, via Wikipedia

Vortex street. Credit: Bob Cahalan, NASA GSFC, via Wikipedia

While this is a very distinctive flow that exists at a specific range of Reynolds numbers, you see flows of all different kinds of Reynold numbers in the real world, too, and not only on my plate. Below, for example, the Reynolds number is higher and the flow downstream of the obstacle distinctly more turbulent than in a vortex street. It’s a little difficult to compare it to the drawing of streamlines above, though, because the standing waves disguise the flow.

2017-09-10-15-16-00

One way to manipulate the Reynolds number to achieve similarity between the real world and a model is to manipulate the viscosity. However that is not an easy task: if you wanted to scale down an ocean basin into a normal-sized tank, you would need fluids to replace the water that don’t even exist in nature in liquid form at reasonable temperatures.

All the more reason to use a large tank! :-)

Why we actually need a large tank — similarity requirements of a hydrodynamic model

Leave a reply

When talking about oceanographic tank experiments that are designed to show features of the real ocean, many people hope for tiny model oceans in a tank, analogous to the landscapes in model train sets. Except even tinier (and cuter), of course, because the ocean is still pretty big and needs to fit in the tank.

What people hardly ever consider, though, is that purely geometrical downscaling cannot work. Consider, for example, surface tension. Is that an important effect when looking at tides in the North Sea? Probably not. If your North Sea was scaled down to a 1 liter beaker, though, would you be able to see the concave surface? You bet. On the other hand, do you expect to see Meddies when running outflow experiments like this one? And even if you saw double diffusion happening in that experiment, would the scales be on scale to those of the real ocean? Obviously not. So clearly, there is a limit of scalability somewhere, and it is possible to determine where that limit is – with which parameters reality and a model behave similarly.

Similarity is achieved when the model conditions fulfill the three different types of similarity:

Geometrical similarity
Objects are called geometrically similar, if one object can be constructed from the other by uniformly scaling it (either shrinking or enlarging). In case of tank experiments, geometrical similarity has to be met for all parts of the experiment, i.e. the scaling factor from real structures/ships/basins/… to model structures/ships/basins/… has to be the same for all elements involved in a specific experiment. This also holds for other parameters like, for example, the elastic deformation of the model.

Kinematic similarity
Velocities are called similar if x, y and z velocity components in the model have the same ratio to each other as in the real application. This means that streamlines in the model and in the real case must be similar.

Dynamic similarity
If both geometrical similarity and kinematic similarity are given, dynamic similarity is achieved. This means that the ratio between different forces in the model is the same as the ratio between different scales in the real application. Forces that are of importance here are for example gravitational forces, surface forces, elastic forces, viscous forces and inertia forces.

Dimensionless numbers can be used to describe systems and check if the three similarities described above are met. In the case of the experiments we talk about here, the Froude number and the Reynolds number are the most important dimensionless numbers. We will talk about each of those individually in future posts, but in a nutshell:

The Froude number is the ratio between inertia and gravity. If model and real world application have the same Froude number, it is ensured that gravitational forces are correctly scaled.

The Reynolds number is the ratio between inertia and viscous forces. If model and real world application have the same Reynolds number, it is ensured that viscous forces are correctly scaled.

To obtain equality of Froude number and Reynolds number for a model with the scale 1:10, the kinematic viscosity of the fluid used to simulate water in the model has to be 3.5×10-8m2/s, several orders of magnitude less than that of water, which is on the order of 1×10-6m2/s.

There are a couple of other dimensionless numbers that can be relevant in other contexts than the kind of tank experiments we are doing here, like for example the Mach number (Ratio between inertia and elastic fluid forces; in our case not very important because the elasticity of water is very small) or the Weber number (the ration between inertia and surface tension forces). In hydrodynamic modeling in shipbuilding, the inclusion of cavitation is also important: The production and immediate destruction of small bubbles when water is subjected to rapid pressure changes, like for example at the propeller of a ship.

It is often impossible to achieve similarity in the strict sense in a model experiment. The further away from similarity the model is relative to the real worlds, the more difficult model results are to interpret with respect to what can be expected in the real world, and the more caution is needed when similar behavior is assumed despite the conditions for it not being met.

This is however not a problem: Tank experiments are still a great way of gaining insights into the physics of the ocean. One just has to design an experiment specifically for the one process one wants to observe, and keep in mind the limitations of each experimental setup as to not draw conclusions about other processes that might not be adequately represented.

So much for today — we will talk about some of the dimensionless numbers mentioned in this post over the next weeks, but I have tried to come up with good examples and keep the theory to a minimum! :-)

Similarity requirements of a hydrodynamic model

1 Reply

Why downscaling only works down to a certain limit

When talking about oceanographic tank experiments that are designed to show features of the real ocean, many people hope for tiny model oceans in a tank, analogous to the landscapes in model train sets. Except even tinier (and cuter), of course, because the ocean is still pretty big and needs to fit in the tank.

What people hardly ever consider, though, is that purely geometrical downscaling cannot work. I’ve talked about surface tension a lot recently. Is that an important effect when looking at tides in the North Sea? Probably not. If your North Sea was scaled down to a 1 liter beaker, though, would you be able to see the concave surface? You bet. On the other hand, do you expect to see Meddies when running outflow experiments like this one? And even if you saw double diffusion happening in that experiment, would the scales be on scale to those of the real ocean? Obviously not. So clearly, there is a limit of scalability somewhere, and it is possible to determine where that limit is – with which parameters reality and a model behave similarly.

IMG_5320

Mediterranean outflow. Mediterranean on the left, Atlantic Ocean on the right. The warm and salty water of the Mediterranean Outflow is dyed red.

I’ve noticed that people start glazing over when I talk about this, so in the future, instead of talking about it, I am going to refer them to this post. So here we go:

Similarity is achieved when the model conditions fulfill the three different types of similarity:

Geometrical similarity
Objects are called geometrically similar, if one object can be constructed from the other by uniformly scaling it (either shrinking or enlarging). In case of tank experiments, geometrical similarity has to be met for all parts of the experiment, i.e. the scaling factor from real structures/ships/basins/… to model structures/ships/basins/… has to be the same for all elements involved in a specific experiment. This also holds for other parameters like, for example, the elastic deformation of the model.

Kinematic similarity
Velocities are called similar if x, y and z velocity components in the model have the same ratio to each other as in the real application. This means that streamlines in the model and in the real case must be similar.

Dynamic similarity
If both geometrical similarity and kinematic similarity are given, dynamic similarity is achieved. This means that the ratio between different forces in the model is the same as the ratio between different scales in the real application. Forces that are of importance here are for example gravitational forces, surface forces, elastic forces, viscous forces and inertia forces.

 

Dimensionless numbers can be used to describe systems and check if the three similarities described above are met. In the case of the experiments presented on my blog, the Froude number and the Reynolds number are the most important dimensionless numbers.

The Froude number is the ratio between inertia and gravity. If model and real world application have the same Froude number, it is ensured that gravitational forces are correctly scaled.

The Reynolds number is the ratio between inertia and viscous forces. If model and real world application have the same Reynolds number, it is ensured that viscous forces are correctly scaled.

To obtain equality of Froude number and Reynolds number for a model with the scale 1:10, the kinematic viscosity of the fluid used to simulate water in the model has to be 3.5×10-8m2/s, several orders of magnitude less than that of water, which is on the order of 1×10-6m2/s.

There are a couple of other dimensionless numbers that can be relevant in other contexts than the kind of tank experiments we are doing here, like for example the Mach number (Ratio between inertia and elastic fluid forces; in our case not very important because the elasticity of water is very small) or the Weber number (the ration between inertia and surface tension forces). In hydrodynamic modeling in shipbuilding, the inclusion of cavitation is also important: The production and immediate destruction of small bubbles when water is subjected to rapid pressure changes, like for example at the propeller of a ship.

It is often impossible to achieve similarity in the strict sense in a model experiment. The further away from similarity the model is relative to the real worlds, the more difficult model results are to interpret with respect to what can be expected in the real world, and the more caution is needed when similar behavior is assumed despite the conditions for it not being met.

This is however not a problem: Tank experiments are still a great way of gaining insights into the physics of the ocean. One just has to design an experiment specifically for the one process one wants to observe, and keep in mind the limitations of each experimental setup as to not draw conclusions about other processes that might not be adequately represented.