Learning Forum

[papp]

Hello, I've recently discovered something quite weird about the heat transfer model of Discovery Live 19.1. The analysed geometry is a straight, cylindrical pipe with a simple heating/mixing section. The fluid is air, and the boundary conditions are the following:

• Mass flow rate in: 1 [kg/s] (at tle left side)
• Temperature at the inflow: 0 [°C]
• Initial temperature: 0 [°C]
• Heat flow in: 1009 [W] (applied on the cylindrical face between the 2 black lines on the left side)
• Pressure: 0 [Pa] (at the outflow)

Using the data of the Discovery Live Material Library, the specific heat of air is considered c = 1009 [J/kg/K]. Based on the definition of specific heat, the formula for temperature rise (shown in the attached image) is obtained , which yields 1 [°C] for the temperature difference between the 2 ends of the pipe using the above parameters. However, Discovery Live produces a significantly larger temperature difference than expected: 43 [°C]. (The mean temperature of air at the outflow area is shown on the chart.) Can you provide any explanation to this problem? Or point out any mistakes in my theory. Thanks, Bálint  

[Naresh Patre]

Hello  Bálint Papp

I will look into it and get back to you.

[papp]

Hello, I have carried out further investigations into this problem. I did a parameter analysis for all inputs separately:

• Heat flow (Q_dot)
• Specific heat of air (c)
• Mass flow rate (m_dot)

The output temperature at the right side was monitored continously. I also registered the multiplier coeff. (C) between the theoretical results (deltaT) and the Discovery Live output (deltaT_DL). The results are the following:

• If the heat flow is multiplied by an arbitrary constant k, the output temperature is going to be k*deltaT. This meets our expectations based on the formula for the specific heat (shown in the original post). Therefore, for this cases, the value of C is nearly constant.
• However, if we change either the specific heat of air or the mass flow rate, the output remains the same temperature, regardless these inpouts' value. This seems to be a serious error of the model.

Thank you for your answer in advance. Bálint

[Naresh Patre]

Hello  Bálint Papp 1. It could be possible that the mass flow rate value specified at inlet could be unrealistic for the geometry in consideration. You can plot the velocity contours and check the values which will give you information about the velocity with which the air is flowing in. You may have to modify the mass flow rate values accordingly to get reasonable velocity values. Please find attached a test case where the theoretical detla_T and DL delta_T difference is much less. 2. You may want to run the simulation with higher Fidelity setting 3. Also, make sure that the output results are stabilized. If needed, revisit the geometry and boundary conditions

🛈 This post originally contained file attachments which have been removed in compliance with the updated Ansys Learning Forum Terms & Conditions

[papp]

Hello Naresh Patre 

1. I will definitely check your model. Mine was a 500 [mm] diameter pipe, which yielded 2 to 32 [m/s] mean velocities at the outlet. I don't think these are close enough to sonic speed to make the model unrealistic. However, if you have a different opinion, please share it.
2. I will be able to run the models on a significantly more powerful GPU in the very near future.
3. What do you mean by the stabilized results exactly? I waited quite a long time for the outputs to settle, and also did a time-integration to smoothen the values, as shown in the figure attached to the original post. Can you suggest anything else, or is this sufficient?

I will get back with the results using the more powerful GPU.

Bálint

[papp]

Hello Naresh Patre I'm back with results. GPU: GTX 1080Ti, Fidelity set to max. As computing capacity is increased (more powerful GPU, higher Fidelity, possibly refined mesh), the numerical results do get closer to the theoretical values. I have used the model you provided for the following study. It has to be noted, that originally the initial temperature in the pipe was 20 [°C] (default data), I have set it to 0 [°C] in order to obtain results more simply. This shouldn't have any influence on the outcome. Experience:

• The temperature change is directly proportional to the heat flow (Q_dot, red curve), as expected and shown by the above experiment. This means, that the calibration of the Discovery Live results can be achieved by a simple multiplication by a constant (C).
• The mass flow rate (m_dot, blue curve) and specific heat of air (c, green curve) are not able to fulfill the desired direct proportionality described by the definition of entropy, as shown in the attached table, the values of the C calibration constant clearly changes as the input parameter varies.

Please, explain these discrepancies (especially the offset of the blue and green trendlines) based on your technical knowledge and expertise regarding the developement of Discovery Live. Thank you for your answer in advance. Bálint    

[Naresh Patre]

Bálint Papp I will look into it and get back to you.

[papp]

Naresh Patre is there any progress?

[Naresh Patre]

Bálint Papp My apologies for missing out on this problem. I will look into it and get back to you on Friday (as I am on vacation tomorrow).

As a side note, it is worth noting that Discovery Live is really not intended to measure the accuracy of results. Rather it is better suited to understand the trend of results due to geometry changes. If you are really interested in measuring the accuracy of results and their comparative study, I would suggest you go for Discovery AIM.

[papp]

Naresh Patre OK, take your time.

I'm aware of the goal of Discovery Live. Its results are to be compared to simulation data from Fluent along with laboratory measurements.

The main purpose of this question/investigation is to test the accuracy and reliability of Discovery Live and to understand its simulation methods, since no detailed documentation is available. Therefore your first-hand explanation on this behaviour regarding heat transfer would be quite useful and much appreciated.

[Naresh Patre]

Hello Bálint Papp 

If you look at the table which you shared, it can be seen that Discovery Live is predicting the trends similar to the trends observed with theoretical calculations. For example, when you are increasing mass flow rate, the calculated delta T is reduced. The same reduction is observed in DL results. If you are increasing heat load, the increase in temperature is observed in both delta T.

The constant [C] that you are deriving from both results is not something which is recommended to check when using Discovery Live. For such comparison, Discovery AIM is an appropriate tool to go for as mentioned earlier. I would request you to perform a similar study in AIM and let us know if you notice any such inconsistency.

The trend results from the table confirm the accuracy and reliability of Discovery Live for which it is primarily designed. So, it is a matter of studying the trends of results rather than the absolute values and not comparing the theoretical values to the simulation values reported by Discovery Live.

Again as said earlier,  Discovery Live is really not intended to measure and compare the accuracy of results. Rather it is better suited to understand the trend of results due to geometry changes, changes in boundary conditions, etc.