Discussion Forum

Hey,

I'm just starting off with this Comsol business and loving it--Version 4.2. Unfortunately, I find myself working in a time crunch and helplessly inexperienced. Here is the problem:

I am modeling a long fuel pipe closed at both ends with an internal fuel pressure of 150 psi. The fuel is hotter than the ground and heat is transferring out of the fluid. I modeled the pipe as being two-dimensional with a set heat flux out of the fluid and initial temperatures of the fuel and the piping. As heat is transferred out of the pipe, the pressure in the tube decreases. The heat transfer creates rings of constant temperature with the center core close to the initial temperature and the outside ring a few degrees cooler. Simple enough.

My questions are...

1. How can I find the average temperature of the fuel in the pipe.
2. How can I graph a pressure v. time graph of the system.

You guys are awesome. Many thanks.

Tim

Hi Tim,

For yours questions:

1. Right click on Definition, model coupling, average. Then, choose the domain or boundary that you want to calculate. So, a function like "intop1" will appear in () on the "Definition branch". You can use this function to estimate the mean value of temperature by typing: intop1(T) in "Expression" field of Global (1D Plot group in Results options)

2. Right click on results, choose 2D plot, right click on that one, you have many options to plot. In the "expression" field, choose your variable by clicking on "2 triangles top left".

Hope that can help U

Regards

Dinh

Many thanks! That was a great help.

Here is my next issue: I am trying to model the pressure drop in the fuel as a result of heat flux out of the fuel.
My current model has a flat line pressure vs. time graph.

I modeled the fuel using a heat transfer in fluids to model the heat flux and resulting temperatures, and a laminar flow that allows me to input pressure. The material is a user defined material with functions for the different properties. The model is attached.

It seems as if the two physics do not interact; there is no pressure change modeled from the laminar flow because there is no velocity field in the closed pipe, and conversely there is no pressure drop predicted from the heat transfer because it is merely looking at heat.

My question...

1. Is there a better way to model this simple problem that would demonstrate the pressure change due to heat transfer? Should I be using a different physics component? Or would I simply need to calculate the pressure on my own from the resulting temperature and fluid properties?

Hi

Here are some comments:
first a trick for your geometry:
in "Finish Union" mode, you do not need to do differences so your 2 last operations can be ignored or simply deleted. In fact the second circle too, but then you should add a layer of size (12.750/12-11.376/12). In the latter case you will get an annulus in 4 segments, these extra internal boundaries are not important, but can ease the meshing, i.e. use a sweep mesh on the external pipe segment, what you cannot do if it's one single full circular part

You should not define 2 physics, as this adds 2 temperatures and make life far to complex. First in the first HT select all, add a Heat transfer in Solids sub node (right click + domain HT solid). Then add an initial value for your solid if you want it different from that of the fluid. Then you have perhaps convecting cooling on the external side of the metal wall add a boundary heat flux or convective cooling (these are the same to a "hidden" sign for the flow direction, check carefully that you remove heat and not add it ;) Then remove HT 3

Then you have the spf but your flow is perpendicular to the 2D image so you do not have any easy way to define inlet outlet etc. Anyhow if its the heat transfer in fluid you are interested in you should use a NITF Conjugated Heat transfer, and then you are better off in 2D-axi to include the tube length. But that is a complex physics and not trivial to get to converge

--
Good luck
Ivar

Thanks for the help. A little of your wisdom can take me a long way.

I updated the fuel pipe model to be 2D axi-symetric. I am looking at a 1 meter long section because the entire pipe would be too long to model. I am using the fluids Non-Isothermal Flow physics component to try and model the heat transfer out of the pipe with stationary fluid. I defined the outside edge of the pipe to be an open boundary so that the heat was not confined by thermal isolation. The model is attached.

I know the total heat transferred out of the fluid in an hour and can calculate the heat flux. I know the heat flux out of the fluid but cannot seem to model it correctly. The solver produces no temperature change on the fluid. Is there a problem with the setup of the physics? Is it possible to define the heat flux at the fluid-wall boundary? Should I use a heat source instead of a heat flux?

Any guidance you can provide would be a great help.

Cheers,

Tim

Hi

I have some problems understanding your model.

First of all if your fluid is immobile, cant you consider simply conduction and now mass transfer?
Then HT is enough and you consider the fluid as "solid" but you do not solve for the elastic properties.

But in your present model your "open boundary" (allows fluid to freely flow in and out (motion) does nothing as it's attached to a "solid"

you might have heat leaving a fluid boundary, then apply an Heat transfer - outflow to allow heat to leave out on that (fluid) boundary, then its not "heat isolated"

Finally you need to do a little homework on the heat diffusivity alpha = k/rho/Cp for steel its about 12[mm^2/s] and for water 0.15[mm^2/s] (I always calculate them as parameter to get a feeling how they behave. basically in 1 sec a heat pulse will penetrate some 12 mm in the steel, but only 150 um in the water, this gives you a good indication how you should mesh for a given time step, so your boundary mesh along the fluid remains of interest, and you would rather need hours than minutes to get any significant T change for your model

--
Good luck
Ivar

Great thanks. I appreciate it.

Modeling the fluid as a solid would work well for just the heat transfer, but I am trying to model the pressure change in the fluid as a result of the heat transfer. Is that possible with the Non-Isothermal Flow physics or do I need to input the temperature-pressure relationship for the fluid on my own? The fluid I am ultimately using has user-defined properties and so I am using water temporarily to work out the physics.


Many thanks,

Tim

Hi

Modelling pressure change implies also stress in the solid (from the thermal expansion). Just as you can consider a fluid (just one liquid assuming it is at rest without convective motion, surrounded y a closed solid volume, a valid hypothesis among others possible) as an expanding material based on it's volumic thermal expansion. With some material property tweaking I believe you can remain in TS, by treating your liquid at rest as a pseudo "solid". But if you have the slightest gas (2 fluid system) in your volume eveything changes. Just as, normally on earth we feel the gravity and that too has an important effect on the pressure within a vessel filled with a liquid.

The main drawback with the CFD and HT is that you also solve for the liquid velocity, and I fair the solver will not converge easily if you try to impose spf.U=0

By adding gravity and boyance forces you will see the convection develop, and that is also what is really happening in your case, but its tricky to solve (requres time and RAM) and I'm not sure it's that you are after, even if it will certainly change quite a lot the heat flow & time ocnstants for thermal exchange within the vessel.

I havent tried all this out, for a similar case, so I cannot say directly go for this or that physics, I would have to try myself and that unfortunately would take too much of my time, even if I find it very interesting. Hopefully someone else has tested these cases and can give us some good advices


--
Good luck
Ivar

Great. Tack för alla era hjälp!!

Hey,

I simplified the model to solve for the temperature change in the fluid after one hour. I used a 2-D model of a cross section of the pipe and imposed a heat flux our of the fluid. From there I used an average model coupling to calculate the average final temperature of the fluid.

The temperature change I expected based on thermodynamic calculations was greater than what the model produced. Thinking the problem was in the material properties of the user-defined fluid, I varied the specific heat. Based on the equation..... Q =( m ) ( Cp ) ( delta T) ..... At a fixed Q determined by the heat flux, the specific heat should be proportional to the change in temperature. However, when I multiplied the curve fit for specific heat by 10, there was no change in the average temperature of the fluid after one hour. What is the reason for this?

Also, how does Comsol calculate the average temperature of the fluid? There are three options for doing so and I am not quite sure the differences. Does it take a weighted average of the temperature at the different segments of the mesh, or does it use another method?

Many Thanks,

Tim

Hi

you model looks still a bit weird for me ;) You have some missing material properties, some undefined materials. And the "outlow" BC is for mass transport boundaries where the mass of liquid passing by takes heat with it, I do not expect that froma firm solid wall. That is heat conduction you have on that boundary, not mass transport heat exchange

I believe you should do a few more examples from the library (try the "cold_water_glass" and the "continuous casting" perhaps start with the "chicken_patties" these examples gives you a better understanding of how to apply the different BC, you need to get used to them and that takes a few models. The best is to print the pdf, to read it and then to follow the instruction. At the end one can compare with the existing prepared model

--
Good luck
Ivar

Hey,

I'll look back through the examples to gain a better understanding of Comsol. The only thing we know about the model, however is the total heat flux out of the fluid, so I modeled it this way to simplify it down to purely looking at the temperature change in the fluid. The gasoline material I used as a reference to make sure my material property curve fits were input correctly and had proper units. On that note...

You mentioned that I am missing some material properties. Looking at other fluids in the materials section, those were the only properties that were provided. What other material properties do I need to accurately represent the fluid? Also, could you explain why increasing or decreasing the curve fit for specific heat by a factor of 10 or 100 produces no change in the average temperature of the fluid after 1 hour? From the equation, Q = m * Cp * DeltaT, if the mass in the pipe does not change, and Q is a constant determined by the overall heat flux, shouldn't temperature change vary inversely with specific heat?

Thanks again for all your help!

Tim

Hi

I connot explain your results, apart it gives me the impression something is wrongly defined in the model.

If you are interested in the fluid exchange and you know the heat loss per area for a given (lets say 1[m]) long tube, then you can take (to start with) a "D HT model, draw a circle of the fluid (forget any surrounding metal), set a T initial to some T0+DT, define a heat flux (negative if its the "inward BC) on your boundary and run a time transient solver.

As you ahve no Heat source the temperature will decrease and eventually fall below 0 K, this is not physical, but consistent with your model.
Then normally the heat loss is given by a heat exchange term in the form a a h[W/m^2/K]*(Tboundary-Tambient) hence the heat source is not constant but depends on the gradient of the local teperature

This you obtain by replacing your HEat Flux BC by a Convective cooling with a reasonable "h[W/m^2/K]" value and a Tambient = T0. Now when you solve you get a more coherent reply as the temperature stabilises to T0 and does become more an more negative.

Then you need to add the heat flux arriving in the "Z" direction from the fluid flow (I assume there is some fluid flow). You can simulate this by saying you have a laminar flow with a parabolic pattern, maximum speed on axis, "0" on the border. Even if your simulation is "static" you can add a domain heat source that you define as the product of the local fluid Z velocity [m/s]* density[kg/m^3] * Heat Capacity[J/kg/K] * (Tin-Tout)[K] = Q_in[W/m^2]. Remains how to estimate (T_in-Tout) which ais the temperature drop (gradient) per meter tube length. That you have probably measured, when you estimated the heat loss

In this static way you can get a first good estimate. You can then improve it by using 2D axi to simulate 1 meter length tubing, witha true heat flow in via the fluid transport, add the material for the tubing solve the fluid velocity and the fluid- structure heat exchange. And compare, you will get somewhat different results, as the (Tin-Tout) I talked about above was constant, as now you will see its in fact rapidly changing with the fluid radius value

--
Good luck
Ivar