Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Consulting a thermal model (soon to be coupled)
Posted Nov 23, 2009, 1:24 p.m. EST 15 Replies
Please login with a confirmed email address before reporting spam
I wish to consult about a 2D heat transfer problem I wish to solve. I want to model the convective heat transfer between a capillary blood vessel and its surrounding tissue. In the image attached you can see the model geometry: the circle is the capillary cross section (I’ve neglected the vessel wall thickness and modeled the capillary as a circle full of blood) and the surrounding tissue. My main interest is blood’s cooling effect on the tissue and not the blood flow behavior.
At first I thought of using the bioheat equation model, but as I intend to couple mechanical strains to the problem, I’ve reached a conclusion that using the general heat transfer model will better, so in the future I could use the same definitions as I use comsol mutiphysics own thermal-structural model.
I wanted to ask:
1. Am I right about the decision of the general heat transfer model? (the main factor in my decision was that I want to solve the mechanical and thermal problems as dependent on one another).
2. Assuming I will expand the model more and wish to add diffusion analysis as well, can I add the diffusion model in comsol to the problem? Will it be a problematic model coupling three analyses? Do I need an equation to couple the thermal effects on the diffusion (I’m interested in mapping diffusion as effected by thermal changes).
3. As my model will be based on a CAD file and as I wish to model all stated above, do you recommend I’ll use a 2D or a 3D model? reading comsol’s guides I’ve noticed that usually when modeling convection a 3D model is used or at least a 2D model in which the axis along which the convective fluid flows (especially, after reading the heat exchanger and thermos examples from heat transfer module user’s guide)… as I’m dealing with a cross section (like in the image attached) I find it quite hard to decide which boundary conditions I better choose.
I know I ask a lot of questions… and will appreciate your help a great deal.
Yael
Attachments:
Please login with a confirmed email address before reporting spam
Well bioheat is no my domain so I cannot give any authorative answer for that one, nor if a standard thermal model is better for you, but its rather quick to change no ? So why not start with what you feel "simpler".
Now concerning adding a third physics, why not, no real limitation there in COMSOL, so long you manage to get the dependent and independent variables names unique or you can ensure by geomentray that there are no naming conflicts.
For the 2D-3D issue, I would start in 2D, as this is much simpler and solves quicker, then at the end I extrude the third dimension to check my 3D case, mostly it is not really necesary, it depends a lot on the physics nd the model you need.
In 2D you must juts continue to think 3D but with a depth (third dimension) of default 1[m] to get your units correct, its that simple.
In anycase good luck an keep us informed on the progress, always something useful to learn fo us other too
Ivar
Please login with a confirmed email address before reporting spam
Thank you Ivar!
You helped me a lot!
However, there are a few things I’m still not sure of and would like to ask:
1. I’ve read in the comsol multiphysics modeling guide, that using the predefined multiphysics couplings for structural mechanics (Plane Strain in 2D) with the General Heat Transfer application mode, the interaction is a one-way coupling using the temperature described by heat transfer application mode to define the strain temperature.it’s also stated there that, By default, COMSOL Multiphysics solves for the temperature and displacements simultaneously.
I just want to make sure: if I choose to couple by myself (add the modules from the model navigator) heat transfer and structural mechanics modes, the two physics will be solved at the same time but there won’t be a coupling like in the comsol predefined multiphysics of structural-thermal interaction, am I right?
Also, just to make sure, in case I want to couple the problems by myself and want to have dependency between the two added applications modes: I need to insert comsol an equation coupling the variables, am I right?
2. As blood has a cooling effect on the tissue, I thought of using the “heat sink” option. “heat sink” can be chosen in both subdomain and in boundary conditions… what’s the difference between choosing heat sink (or source) in the subdomain or in the boundary? I’m not sure which one I should choose as the thermal convection between blood and tissue takes place on the vessel’s boundary but itself has a metabolic heat…
3. Also, in the subdoamin definitions I can define velocity fields components in the x and y directions… If I understand correctly, this is the velocity of the fluid causeing the thermal convection (am I right?)
As blood in my model flows in the direction coming out from the page (screen) ,as you can see in the image I’ve attached in my previous post, I’m not sure x or y direction are suitable… what direction can I model than?
Your answers are most appreciated,
Yael
Please login with a confirmed email address before reporting spam
I'm not sure I getting all what you ask, but I'll try to reply at best, and from what I know, as there is still many items I'm neither 100% sure about in COMSOL.
The predefined application modes hare mainly an addition of two or more physics, with the individual settings adapted, and when possible, groups already made and named, possibly few ODE added as well as scalars etc.
To understand the differences, I mostly load the individual application modes (physics) and then I add the complex application mode.
I define a simple subdomain, a few BC (boundary conditions) so that the solution solves to something not fully trivial so I may run it later.
Then I systematically go through all items in the model tree (on the left), and see how the two elements differ, w.r.t. sub-window settings, equations, scalar variables etc
Basically I notice that the cobined heat structural has the temperature tick boks in the structural subdomain setting already ticked on with the temperatre variable name of the heat transfer mode. And you might need the pressure of the structural to defined some optional pressure variables in the thermal BCs
So long the temperature of the heat transfer and the pressure of the structural physics is not linked by some physics I do not believe there is much more (but pls check carefully).
2)
Source /sink body versus surface: mainly for me there are questions of how and where the souce/sink is generated. In the body=subdomain its uniformely distributed in the full volume, from the boundaries it will diffuse into the volume and establish an equilibrium, probably you need a combination of both.
Now with the thermal physics you can either have a volume of some material, or just a boundary exchanging with ambiant air (or somethin else) and you define "just" the local condition on one side of the boundary.
3)
the velocity feld in the thermal Convection tab is the motion of the "fluid", as there will be mass transfer this will renew the volume in contact and the heat exchange might vary. The question now is how this velocity changes, is it a fixed external velocity, or is it driven internally by some physics, such as a gravity field and an ideal gas law ? here the pressure pa might (or might not) be linked to the structural pressure that will depend on your model.
But I beleive for youre case the "pa" of the blood fluid should apply to the surfaces of the vessel interiour, hence onto the structure
Now when you are in 2D you should think of it as if in 3D but with a uniform tickness of 1[m] in the depth of the screen/paper, all formulas and physics must be slightly adapted, but it's easy: check your units!
COMSOL has done an excellent job there to show how important it is to use consistent units, and how easi it is to check what you are doing !
Hoipe this helps
Good luck and keep us informed on the progress
Ivar
Please login with a confirmed email address before reporting spam
Regarding your sentence: “So long the temperature of the heat transfer and the pressure of the structural physics is not linked by some physics I do not believe there is much more (but pls check carefully).” By saying linked by “linked by some physics” do you mean by an equation? And if so, do I need to insert comsol such an equation?
In the structural definitions I have defined a normal load on the blood vessel boundary to represent the blood pressure (the load tab in the boundary settings) … but I don’t understand how to model the blood velocity as in the heat transfer model definitions the velocity field in the thermal Convection tab is given in the x,y direction and I need the direction coming out of the screen… I don’t understand how to model such a velocity directions… also, I don’t see how in comsol are the pressure (from the structural model) and velocity (from heat transfer model) are connected to one another…
Well, I have come up with one more question, as I wish to solve the problem in time (I want to define strains and see how it affects the shape and the temperature field after 2 hours the model is loaded by the strains defined). I’ve defined a constant strain on the upper boundary (constant in time). When I solve in transient state the “time solver” remains in 0%... and the model isn’t solved… I get a solution only when using the static solver… what is the problem with the time dependent solver? Must I define a time dependent strain?
Also, I don’t see in the solution I receive that the thermal map was affected by the strains… I’m afraid the strains and temperature fields are not really coupled and are only solved simultaneously …
Thank you again, looking forward to read your answers,
Yael.
Please login with a confirmed email address before reporting spam
the velocity feld in the thermal Convection tab is the motion of the "fluid", as there will be mass transfer this will renew the volume in contact and the heat exchange might vary. The question now is how this velocity changes, is it a fixed external velocity, or is it driven internally by some physics, such as a gravity field and an ideal gas law ? here the pressure pa might (or might not) be linked to the structural pressure that will depend on your model.
But I beleive for youre case the "pa" of the blood fluid should apply to the surfaces of the vessel interiour, hence onto the structure
Now when you are in 2D you should think of it as if in 3D but with a uniform tickness of 1[m] in the depth of the screen/paper, all formulas and physics must be slightly adapted, but it's easy: check your units!
COMSOL has done an excellent job there to show how important it is to use consistent units, and how easi it is to check what you are doing !
Thank you very much for your help. However, I do have problems trying to model the blood pressure...
I wish to model a blood pressure of 4266 Pa that will be applied on the surfaces of the vessel interior boundaries. Using the Load tab in the boundary settings of the structural mechanics subdomain settings I can only define a load which is force/length or force/(length*thickness). I know the length and thickness in my model but I can’t translate the blood pressure from Pa into N as blood pressure in the human body can not be translated in that way… how can I define a pressure equal to 4266 Pa on the vessel interiour wall without using a fluid model also? I don’t wish to use the fluid application (or Navier stokes) as I don’t want to model the fluid flow (velocity).
I've heard about using a PDE for boundaries... but I don't understand which term can be applied as pressure and in what units...
Help is most appreciated,
Yael
Please login with a confirmed email address before reporting spam
You have many questions there; and I do not fully understand all, here is something to think over:
Let’s take first the pressure case. Why can’t you use the 4266[Pa=N/m^2] as a static value ?
Blood being a fluid (basically water) and rather incompressible it will distribute the force on its adjacent vessel walls in a normal way giving a surface pressure of 4266[N/m^2], which for a "2D" and by default 1[m] deep model means 4266[N/m] per "dl[m]" line element. Once integrated you get a "line(2D)=surface(3D) area on which you apply a FORCE in [N/m] where the [1/m] is for the per 1[m] depth.
As I said before, look at a 2D model as a 3D model, with a default 1[m] depth.
Then forces in [N] become [N/m] to cope for the per 1[m] depth normalisation.
The only thing to bother about is the normalisation of the integration, if you model the blood vessel section as a circle of radius "r[m]", then the perimeter is "2*pi*r[m]" and the total force is "(4266[N/m^2]*2*pi*r[m])[N/m]" on the full circle.
You can sole for "initial conditions" and check that the applied force on the surface gets to this value in [N/m] (per[m] depth), and you can have a boundary line integration to get the total blood vessel perimeter calculated automatically for you.
Nothing different from blowing up a balloon. If you put an internal pressure or 1[atm] (in addition to ambient) you have an internal pressure of some 1E5[Pa]=1E5[N/m^2]=10[N/cm^2] or about 1[kgf/cm^2], the total absolute force you apply normal to the balloon shell will depend and scale proportionally with the balloon area, which is directly related to the square of the balloon radius. If you now say that your balloon has the shape of a cylinder the area is the depth "w" times the cylinder perimeter Area[m^2] = 2*pi*r[m]*w[m].
So in 2D when defining a surface(3D)=line(2D) pressure load from a fluid, hence applied normal to the boundary, you should:
1) Select "Edge load is defined as force/area using the thickness" (and leave the thickness at 1[m] by default (or whatever you believe is relevant)
2) Select "Distributed load" & "local tangent and normal coordinate system"
3) set Fn (normal "force") to the desired pressure value in [N/m^2], leave Ft tangeant force at "0"
If you do want to use the "Edge load is defined by force/length" you must divide the pressure by the total length of the perimeter, that you can calculate by a “Boundary Edge integration”. Notice that the units changes by selecting this option.
Furthermore, you could select "type of load: "Follower" and use the pressure. This is normally used for large displacements and tells the programme to recalculate the normal direction for each step, as with large displacements the surface normal might change. In a normal "linear" "small deformation" case the "normal" direction is calculated from the initial geometry and is assumed fixed.
In this way you can forget the "fluid" and use only its effect: the pressure on the vessel surface for your simulation (always 1 physics less).
Hope you follow me so far,
Next a few words structural and thermal coupling. The simplest coupling considered is the thermal expansion, as the temperature changes the material swells and this volume increase can affect the stress and boundary pressures, if these boundaries are constrained and not free to move. In most of these cases you can solve for the thermal in static alone, and then separately solve the structural case, this means using segregated solver steps. Or if the model is simple a direct global calculation can be performed, it's just more RAM hungry, and could be less stable). In the structural module, the thermal expansion is included in the physics which means it need to know the temperature at each node, but not how to calculate the temperature fields at these nodes, for that you must add a ht (heat transfer, or bio heat, or electro-heat...) and solve this(ese) physic(s) along the line.
If you want to apply the physics yourself, take first a look on how COMSOL is doing it i.e. load a new model with 2D "multiphysics" Plane Strain (pn) and add the structural module Plain strain (smpn). Then create a volume in there (just select a default 1m^2 square) and then look at the variables COMSOL has defined: "Physics, Equation Systems, Subdomains Settings - Variables" tab. Use the corresponding doc to get more info, but you see here all variables used, their definition and their units.
If you now go into the SMPN "Subdomain Settings, Load - Include Thermal expansion" you should notice that the equations are updated, and in paticular the variable "p" pressure has got a temperature "Temp_smpn" variable, this should be set to "T" from the other "ht" application. It is already set up so if you select the thermal-structure interaction model (=SMPN+HT) in the structural module. Note that there is no variable "T" in the pn or smpn module, in the snmp its called Temp_snmp to avoid conflicts, it’s up to you to "equal" the Temp_smpn=T by adding "T" to the right subdomain entry ("Subdomain Settings, Load - Include Thermal expansion Temp : Strain Temperature"
a few other points:
You probably do not need to calculate the blood velocity, apart that you need to know it if you say that the blood is bringing in heat this is provided at a certain rate, depending on the volume of blood flowing through, this means for a fixed vessel section, the volume of the blood is the section area times the velocity times the time considered =[m^3]
Heat sink/source in subdomain or surface: sub-domain means that all the volume (surface in 2D) is heating/cooling, while boundary/surface means that heat is brought in or out only through some of the boundaries (lines in 2D) i.e. the surfaces might cool-down / heat-up but the temperature inside of your flesh will change slower and you will see gradients as the heat flow from the full volume/area towards the surfaces, and this takes time, so a transient analysis is often very sensitive to such a choice.
Transient / static stress build-up. Often one has a model in a particular state, already with stress applied, temperature having reached a steady state different from 0[K], and then one want to apply some local effect or time dependence. For this either you know everything and you set up all the initial conditions straight and you solve, or you use the software to calculate the initial conditions, you save the solution and your "restart" from here, either manually in two steps, or you use the Solver manager "Sequence" tab so that this can easily be repeated.
Doing multiphysics calculations obliges you to go into and study the individual physics, and to understand how they interact, and how COMSOL is set up to get them to interact.
Once you get acquainted to COMSOL "way" it really becomes fun
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Thank you so much for your thorough and detailed answer.
I appreciate your help so much.
I have indeed defined "Edge load is defined as force/area using the thickness". I’ve defined the thickness of the model to be 150 micron as I think it fits better with my model dimensions (160*120 micron^2).
I have also used the definitions you stated of "local tangent and normal coordinate system" and defined Fn and set Ft=0, but I didn’t see where can I define "Distributed load"? I’m using the load tab in the boundary settings…(comsol 3.3)
Also, Where can I define "type of load: "Follower" ? (as you recommended in your reply) I didn’t know this and this is a very important issue.
Finally, just to make sure: I set Fn to be -4300… I’ve tested the minus sign on a separated structural model to make sure this is a pressure to the outwards direction of the boundaries (making sure the vessel walls won’t collapse).
Currently, I have coupled the plain strain and the bioheat equation, I am initially solving for steady state for the structural and bioheat modulus, and than solving only for the bioheat modulu in transient analysis.
Regarding the convective heat transfer of blood, I will soon a new response to your reply in my post: Boundary conditions in heat transfer (www.comsol.com/community/forums/general/thread/2291/#p5846).
I haven’t posted it yet as I am still checking several things…
Best regards (and thanks again),
Yael
Please login with a confirmed email address before reporting spam
well I have v3.5a, so there might well be some differences there, do no longer have a 3.3 up running, cannot help a lot on those issues, but start like that if you have a "normal" load its already a good start.
For the sign, normally "normal" is going out of the volume, so if the inside of the blod vessel is not part of the meshed geometry, or is inactive the edge normal should go that way and you need a negative sign.
But to be 100% sure, you should check on a simple case for ALL 4 quadrants of the circle, because, I have a doubt just now, in 2D the "normal" could be defined w.r.t. the sign of "ds" and that is given by the direction of the arrow in edge view mode. Just for electric currents along an edge, you must look at the arrow direction defining the +, so for a circle the loop has 4 quadrants with different sign for each one. But this might not apply to force normal, cannot remember pls verify.
In any case have fun
Ivar
Please login with a confirmed email address before reporting spam
I’m trying to use the material model library in order to model the blood inside the capillary blood vessel (until now, I’ve modeled the capillary as a hole which had a load [pa] on the boundary to represent blood pressure).
For start, I’ve decided to use the water (liquid) from the liquid library. After doing that, I intend to add blood as a material to the fluid material library.
In the subdomain settings I’ve chosen water and than defined heperelastic material (I’m not sure about that but I thought it’s the best from the options there…) , nearly incompressible material , and set values for initial shear and bulk modulus…
In the boundaries settings I’m defining a load of 4300Pa (Fn=-4300). My model is solved, but when I’m plotting the pressure, I get that in the capillary I have a pressure of about 582.7 pa… what am I doing wrong?
My model is 160 ,icron in length, 120 micron un height and I’ve defined a thickness of 150 micron…
Also, I’ve noticed that in the subdomain settings I can define pressure in the init tab, should I use that to model the blood pressure as well? I’ve tried to use that… but my model doesn’t converge than… I receive the error:
Error:
Failed to find a solution:
No convergence, even when using the minimum damping factor.
Returned solution has not converged.
I must say that I’ve tried to use the init tab alone, and tried to use it while modeling the boundary load simultaneously… and in both cases received the error…
Should I use the init tab for pressure?
Finally, when I decide to add blood to the fluid material library, Comsol will aggnolidge it as a fluid , right?
Thanks,
Yael
Please login with a confirmed email address before reporting spam
If I read you right you are saying you use water instead of blood, that I can follow (I belive blood is about 92% water, pls check), as a first approximation it's a decent start, then you choose/state hyperelastic, there I do not follow.
2 recommendations:
1a) when using the material library be sure you use the right material a given material can appear several times, and that the one choosen has all the required properties, use the show "all" tab, and select "hide undefined properties", for me these should be the default mode (COMSOL developpers have decided otherwise), then verify that you have chosen the correct phase (i.e. liquid, gas, solid ...) and orientation if applicable.
1b) Then check that the proerties are there, and that they are valid within the desired T and p (temperature and pressure) domain you intend to us for your model (or watever other variable).
1c) select OK and when back in the subdomain window select the appropriate material setting and be sure you have all required variables IN BOLD, as BOLD means this is read-in from the library (you might want to override some values or to complete some other missing, this will appear as normal text), check the units if OK.
1d) Then be sure that all new variables (required by the material data) exist and are defined in your model, such as T, p etc (for some Tempref too), if not set them as constants, "red" units is typically an alert about such a case, I mostly start all my models with 3 constants predefined:
T 20[degC] "ambient temperature"
p 1[atm] "ambient pressure"
G 1[lbf/lb] "gravity acceleration"
2) when you select a material property, be sure you fully understand it, know its definition and applicability, if you have any doubt, have a look at the documentation, i.e. search for hyperelastic in the smeUG.pdf doc, i.e. p120 V3.5a you will see a sentence such as:
"...material models for modeling nearly incompressible hyperelastic materials such as rubber...."
so I would not use hyperelastic for water, anyhow if you look at the Material properties, you see several are in normal text, and are left at default values i.e those of steel or copper or... depending on the application mode selected. If you are interested in the thermal properties of water, I would be sure I have at least heat capacity, heat conduction and density in BOLD, for fluidics, you need in addition the viscuosity.
Furthermore, water as a fluid canot be modelled correctly with the structural application modes such as the smps, smpn ..., you need the Navier-Stokes application ch... for the fluid, and then link this to the structural parts, or you must rewrite some of the equations in there to make them "fluid compatible", the latter would anyhow not be the way I would recommend.
Now for your case, in 2D do you really need the water meshed as a volume ?
I would say no, at least to start with, leave an empty cavity for your "big" blood vessel(s), and apply the pressure directly on the vessel valls/boundaries, where you also model the heat exchange via "h" in [W/m^2/K] (correcting then the units for the default 1[m] depth of a 2D model), and the temperaure of the blod, being the Tinf variable, or via Q and the bioheat equation for a blood vessel boundary or something like
h[W/m^2/K]=rho[kg/m^3]*C[J/kg/K]*v[m/s]. where rho, C and v are the blood density, heat capacity and velocity.
To find reasonable "h" and or blod flow values you will have to look it up, or I can recommend that you get hand on the recent book of Ashim Datta: "Modelling Transport Processes, Applications to Biomedical Systems", you will find most relevant material data and more therein.
Note: use the standard bioheat equation for the irrigated muscle subdomain
As you say, you have also a "p" and "T" (in fact NOT "T" but "Temp_smpn" and "Tempref_smpn" !) that can be defined, for the sme application modes (SMPN as I'm, describing here above), by the "Subomain Settings Load tab" and then selecting, "include initial pressure", and "include thermal expansion", which will add a few physical elements, such as pressure depending on the thermal expansion etc
To understand what this means: look at the changes of "Variables" tab under "Physics - Subdomain settings" this is where COMSOL updates the equations governing the physics for the application mode, depending what you add in the GUI's.
Note also that if you change these load settings, you might have to reapply your library material, as this changes some of the names, check the BOLD text on the Material, if its still there !).
This seems perhaps complex, but once you think really over it, it's rather logical, if you add/replace some new physics and variable names, you might have to update other parts if your model too, so as usual with a nice piece of hardware or software: do not play with the buttons if you do not really know what thay are for;)
or rather, read and re-read the books first.
Finally I agree, I talked about hyperleastic material earlier, but I was referring to the muscle/fat that for me is more like rubber, not the blood, sorry for that confusion. Now if you want to see, in addition to the temperature distribution a compression of the muscle due to the blood pressure from the main blood vessel you need the material properties for muscle/fat, again there I'm not sure where to find those. If you define them as hyperelastic you need the appropriate parameters, I'm not sure finally if this proposal was the best, but I neither know any equivalent E, nu for muscle.
Perhaps we should give google a try
Hope this helps, a little on the way
Ivar
Please login with a confirmed email address before reporting spam
the thread is getting long, rereading number 1 up there I notice that I was refering just above to the SMP.. modules, while you probably are not using that one at all, only the bioheat (htbh) module which should be enough to do a thermal analysis.
I got confused with the refernce to hyperelastic which is structural material property as well as pressure which again leads me to structural or fluidics. If you want to see the effect of the pressure variation of the blood flow, you need the structural model too for the muscle.
Apart from that, most wat I said earlier should be still relevant for you.
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Thank you once again for your detailed reply! You are of great help to me! ?
Reading your answer I have realized I haven’t given enough details, I will try to correct this.
I’m using plane strain (smpn) and bioheat (htbh) modulus. My muscle is defined as a hyperelastic neo-hookean material, with the values of initial shear modulo, bulk modulus, etc… as I find in literature. In the bioheat modulo I have also defined all the properties of the muscle and blood (heat capacity, density, etc…) as found in literature. The convection coefficient was indeed evaluated in the way you have mentioned.
I must say I was rather disappointed to read in your reply that “, water as a fluid canot be modelled correctly with the structural application” and that I need Navier-Stokes application… I don’t want to use Navier-Stokes application because I don’t want to define blood velocity… I thought that by using the model library I can select a material that will fit well with the modulu I am currently using…
Well, indeed at first I’ve modeled my capillaries (which are in the diameter of 10 microns) as a hole in the tissue, and applies a load on the boundary to represent blood pressure. I must add that I’m applying displacement on the upper boundary of my muscle (upper boundary of the rectangle) , as I want to see the deformation. At first I’m modeling a displacement in the x direction; the next step will be to add a y displacement also.
The reason I wanted to add a material to the capillary “hole” was:
1. I thought this will be more exact, then I could also see the stress distribution in the capillary subdomain.
2. as I am coupling the bioheat modulo as well, solving for the thermal map I get the temperature in the muscle, but the capillary still appear as a hole.
However, those matters are not critical issues, just thought it will be nice to have… ?
I know that blood (or water) is not a hyperelastic material (blood is viscoelastic). But after I’ve added water from the fluid material library (I have done the 1a-1d steps you suggested, properties defined in the library for water as liquid are: heat capacity, dynamic and kinematic viscosity, density and thermal conductivity, clicking on the “function” button I have seen that the properties are dependant on temperature and they fit my model) I’ve noticed that the subdomain settings for the material model weren’t changed (besides the density which was taken from the model library and appeared in bold), the material was selected by comsol as default as isotropic with default values of E, etc… as were already in there before the loading of water material from the library… this is why I thought of changing those values and settings…
Well, I guess that if I need to add the Navier-Stokes application, I’ll better leave my capillaries as holes and use the structural modulo alone…
Best,
Yael.
Please login with a confirmed email address before reporting spam
Well then I hadnt forgotten everything, good.
Now were I do not agree is when you propose:
"1. I thought this will be more exact"
Using physics on a boundary is not "less exact", provided that your physics is correct and describe all (most)properties involved.
An incompressible liquid as water has no shape, hence whatever is applied on one wall containing some water is distributed uniformly on the others, your vessel section will not change except due to the thermal expansion difference of water from a temperature change, but for a body these temperature changes are minime and the thermal expansion of water is small.
At least for me pressure on the wall is plenty to describe what happens in the blood vessel in 2D.
Once you go further into the details, one would need a better 3D model with the fluid and its velocity along Z, then the visuosity of the blood would also be needed, and normally the velocity distribution of the fluid in the vessel, which is not uniform, rather parabolic with about 0 velocity strictly on the borders.
This means on the other hand that the formula for h=rho*C*v is supposing a perfect heat transfer wall-liquide and is a "best" case.
The corrective factor for "h" would also depend somewhat on the vessel diameter, the heat conduction and the transverse velocity of the liquide (still neglecting turbulence), as if the conduction is high, the transverse velocity low and the diameter small, the temperature gradient across the liquide can be neglected, and we have a better heat transfer from wall to liquide.
A radial scaling value "a" can be obtained from:
a[m]=k[W/m/K] / (Cp[J/kg/K] * rho[kg/m^3] * vz[m/s])
where k is the heat conduction, Cp the heat capacity, rho the density and vz the transverse velocity of the liquide
Finally, the bioheat equation is basically considering all this by assuming an uniform capillary density for a given volume of body material.
Further, when you go one step further into the details, one would see Venturi effects from the blood flow and its pressure variations of the hart beats, fighting against capilary collaps, but for this I believe you need a 3D model, and such 3D models are very heavy: hours to solve and dozens of GB RAM required.
One way out would perhaps be to push further 3D flow physics on the 2D edge boundary.
In conclusion quite some physics from many domain to apply, true "multi-physics" here
Have fun
Ivar
PS from next week on I'm back at work, so I will have far less time to follow the forum
Please login with a confirmed email address before reporting spam
Guess I better stick to 2D with the load on the capillary boundaries… :)
Thank you again for your help, I appreciate it so much.
Hope you are enjoining your time away from work,
Happy New Year,
Yael
Please login with a confirmed email address before reporting spam
Indeed, Happy New Year,
Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.