Transient heat transfer simulation help: Mug of water (beginner)

[anindo]

Hi, 
I am new to Autodesk Nastran and wanted to learn the ropes by setting up a transient thermal heat transfer.
Problem statement: Hot water at 80 C in a room-temperature stainless steel mug. Ambient temperature is 25 C. I would like to plot the pattern in which the mug and its handle will heat up over time, and then how it will cool. 

Version of Inventor and Nastran: 2025.1

Evidently, I must have done quite a few things wrong, as this plot at 3600 seconds shows: The bottom of the mug is at nearly 80 degrees after an hour, and the rim remains cool.

To begin with: 

1. Modeled a mug with a body of water in it, separate solid bodies in contact. Assigned materials SS and water. Brought into Nastran environment.
2. Auto contacts did not work (pack & go attached for reference), it seemed to work, but on running the sim, it failed with errors that contact was not found at each of the 3 contact faces.
3. Added instead a Solver contact with 0.25mm [EDIT: Changed to 3mm now] active distance and 0mm penetration. No further warnings on that front.
4. Added a Temperature load of 353 degrees (~80 C) to "body<1>@water" with the intent of initial temperature of water at the start of the simulation. Perhaps this is not the correct method.
5. Added convection loads on side faces, rim and handle of the mug but not the bottom, using a mid-range looked up value for stainless steel free air convection, of 0.025 mW/mm^2 K.
6. Added convection load to the top face of the water, 0.05 mW/mm^2 K, again a looked-up mid-range value for water in free air.
   
7. When attempting to run the simulation, it asked for an initial condition, which I assumed is the initial ambient temperature. Hence added a fourth load, an initial condition of temperature 298 K (~25 C), the ambient temperature.
8. Did not add any radiation loads, as those will be minor contributors given the small temperature difference - and also as I found forum posts about convection and radiation not able to coexist on the same faces.
9. Set the time step to be 60 seconds, 60 steps, i.e. 3600 seconds duration. Step method set to CONSTANT.
10. Meshed the model and ran the simulation. The result which is the first screenshot in this post, indicates I got one or more things wrong. 

As I haven't found any basic tutorial for this classic simulation problem of hot water in a mug, I have made many assumptions to get to this point. Please help me improve my understanding, and cool my mug in a realistic duration. 

I would also appreciate pointers to any tutorials for the basics of nonlinear transient heat transfer analysis. 

Thank you!

UPDATE: 
I've fixed a small error in the CAD model, and changed the contact solver active distance. Also updated the loads (see updated attached pack&go file) and TimeStep. This time the simulation seems to have improved somewhat in that the rim of the mug gets hot. Still not solved the main issue, of the expected cooling down of the entire system after a long simulated duration: 

Thanks for your patience and assistance.

[John_Holtz · Accepted solution]

Hi @anindo . Welcome to the Inventor Nastran forum.

The main problem is the temperature load is a load that forces the temperature to remain constant for eternity. That is not what you want.

The message about missing the initial conditions is saying that you need to specify the temperature of the model at time 0 (not the ambient temperature). It is the initial temperature that indicates the water is at 80C and the mug is at 25C at time 0. To have different initial temperatures, you need to follow the steps in this article: How to set up a transient heat transfer model with different initial part temperatures in Nastran

John

[anindo]

Hi John, 

I digested “How to set up a transient heat transfer model with different initial part temperatures in Nastran” down to my specific use case: Ambient temperature 25°C, Stainless Steel mug at ambient, hot water at 85°C filled into mug at time 0, analyze the cooling of the system over time. I also wrote a Python script (CC0 license, attached) to automate adding the TEMP cards from the Linear Static NASTRAN file into the final Nonlinear Transient Thermal Analysis NASTRAN file. Also attached is my process note, should anyone else wish to examine it. 

While the behaviour mentioned in my original post seems to have changed, I still do not get the expected results for the cooling down of a mug of water: Even at the 1-hour timestep, the water temperature is far higher than it would be in the real world. I would appreciate any help with this. 

Here are the results at 30 seconds, 10 minutes and 60 minutes:

• Internal conduction within the water behaves as though in a solid. Fluids will have internal convection and movement causing rapid equalization of temperature within, is there a way to incorporate that in the analysis?
• I looked up the convection coefficient to be between 0.005 and 0.025 mW/mm^2 · K for metals in free air convection, and 0.05 to 1.0 mW/mm^2 · K for water in free air convection. Not sure this is correct.
• “Initial Condition”, subtype “Temperature” is set to 298.16K i.e. 25°C. Not sure this is correct either.

Attaching an updated Pack & Go file in case someone can figure out what else I am doing wrong. 

Regards

--

Anindo Ghosh

[John_Holtz]

Hi Anindo,

• The water is solid because Nastran only has solids. If you want to calculate the transport of heat due to the fluid motion, you need to use Autodesk CFD. The alternative is to increase the thermal conductivity of the "water" to keep the bulk temperature more uniform; that is, simulate the transport of heat that occurs by motion by changing the conductivity. 
• Those convection coefficients seem reasonable for the solid to air. I did not know that water to air had a higher convection, so that is good to know.
• I'm not sure that the initial temperature matters since you are replacing the initial temperatures with your Python code. Or maybe you are only replacing some of the temperatures (either the water or the cup) and using the initial temperature for the ones not replaced (either the cup or the water). As long as the temperature on the first step is correct, you set it up properly.

John

[anindo]

Thank you.

Regarding "The alternative is to increase the thermal conductivity of the "water" to keep the bulk temperature more uniform":
Is there some reference for an appropriate value for still water? Also, I assume this pseudo-conductivity value will have a significant temperature dependence.

So where do I start?

[John_Holtz]

Hi,

I have not researched how to change the thermal conductivity to approximate a real fluid. Your suggestion that the conductivity changes with temperature is reasonable, but it also implies you are looking for a more accurate result than what you will get by adjusting the conductivity. Are you sure you do not need to use CFD (Computational Fluid Dynamics) to calculate what you want?

Perhaps other readers will have some suggestions. 

The alternative is to buy a thermometer to detect when the coffee is too cold to drink, rather than calculate how long you have to enjoy the cup before it gets cold. Ha ha.

John

[anindo]

Hi John, 

Yes, I do eventually want a more accurate result than possible with solid model thermal analytics: One of the end goals is estimating liquid cooling flow rates for LED array stadium lights. However, this isn't a bad start, and I thank you for your insights. 

I will work on buying the Autodesk CFD subscription - and then learning to use it. Meanwhile, NASTRAN will still be useful for solid-on-solid thermal analytics, so it's all good. 

Jokes apart, the thermometer suggestion has merit: I already have a couple of bead probe K-Type sensors in my shopping basket. I'll need those to determine what conductivity value for water comes closest to reality. 🙂

Which of your posts should I mark as "Accepted answer"?

Regards

--

Anindo

[John_Holtz]

Hi Anindo,

I want to make sure I understand your statement "liquid cooling flow rates for LED array stadium lights" compared to your coffee cup analysis.

• "cooling flow rate" sounds like fluid being actively pumped through the system. Where ever the liquid is flowing, that would cause a forced convection on the model which can be calculated in Nastran by assigning a convection coefficient. The temperature rise of the fluid can be calculated by hand since the amount of heat going into the fluid is given by the Nastran analysis, and Q=(mass flow rate)*(specific heat)*(temperature rise).
• In the coffee cup example, the liquid is not actively pumped, and therefore the buoyancy of the fluid (natural convection) becomes an important factor. (If you stir the coffee, then the temperature of the coffee could be considered uniform, and changing the conductivity to keep a uniform temperature would be the way to simulate the "stirring".)
• So back to the real application, is the cooling fluid sitting somewhere in a tank and being cooled like the coffee example? The heat conducts to the tank (and maybe some fins to increase the effectiveness) and the buoyance effects of the fluid are important. Or maybe the LED and cooling are all self-contained so the cooling is more passive than actively pumped.

Regarding which post to mark as the answer, I would suggest the reply where I suggested the article to specify different initial temperatures for different parts. That was the main cause of the original unexpected temperatures (and the temperature load which is also in the same post). The details about the temperature profile in the fluid is a secondary issue.

John

[anindo]

Hi John, 

Two operating conditions: 

1. Normal operation: LED board mounted on a shallow heat exchanger (likely with fins), forced coolant convection. Several options for pumping the fluid, I should be able to pick one based on what flow rate is needed for an acceptable steady-state temperature. This will be an actively pumped liquid cooling analysis as you pointed out - I have found some studies that should help me, but pointers to nice examples are welcome!
2. Failure during operation (typically due to panel or pump failure), needing rapid replacement: How soon will the module be cool enough to pull out of position to swap in another module? This one is the coffee cup analogue.

I am sure I will face many more stumbling blocks as I progress, I hope you're OK with many more queries!

Regards

--

Anindo