Rotating equipment: Which Torque prediction is correct?

[Anonymous]

Hi to all,

I got this question open for a Long time, now... I got rotating Equipments to simulate. I got an impeller, a rotating Zone.

When I check the results during run i see in the GUI a value for the torque prediction for impeller an rotor, additionally in the torque file, which I use for post processing another result for torque. Unfortunately, the three results does not really fit....

So, which result is now correct, why is there such a Deviation and how to fix it?

I normally check the results during run manually and then later on, I "calibrate" the torque results written as file with this Deviation.

Additionally, what I found out: If the Center Point is not 100% complying the cartesian coordinate System, it can happen that the results are complete out of the line...meaning the Deviation between the three different torque values climbs up...

Would be interesting, how to deal with it and find a solution...

Thnx in advance,

kippo

[matt.bemis]

Hi @Anonymous,

Nice to hear from you. I've heard you mention this deviation in torque before, but have not personally really chased this. Are you specifically asking for the relationship between hydraulic force for in X, Y, and Z vs. "hydraulic torque"? In practice I've always pretty much neglected the hydraulic force data because it does not have a lot of use and instead relied on hydraulic torque. I understand why it may be good to understand though. 

Can you give an example of the three results not fitting? You may be speaking about non physical hydraulic forces? 

[Anonymous]

Hi Matt,

well... This is elementary for hydrodynamic machines... To be able to decide which design is more efficient.

The results are standard for all simulations I provided with ADSK CFD.

the Point of gravity is set to 0,0,0.

I defined the rotating Region as volume Group " Rotor".

I try to put some Images in the upload, have a look.

You can see the Deviation between the reults in the GUI during run and the torque results to the same time step in the torque file. This causes a great deviation in the prediction.

I got another case running, where the Volute outlet (diffuser) is not aligned with teh cartesian coordinate System.  then the discrepance between the GUI results and the trq-file is getting larger and I cannot find out which value should be correct.

kippo

[Anonymous]

Here another case:

A big pump.

the rotating Region is proposed in the GUI (during run) with

Torque, 445.152, N-m

the impeller surface (as volume selected) proposes in the GUI

Torque, 484.666, N-m

and in the trq-file a value of:

Hydraulic Torque -4.614707e+02 N-m

the rotating Point is at 0,0,0 and the direction is also correctly selected (besides the minus)...

This causes an Efficiency Deviation of more than 10% in the calculation...

At the Moment I can only trust and hope that the lowest value is teh best (meaning best Efficiency value) and then during post-processing the torque values are lowered in the Excel file.

Also the Deviation is not constant for different flow rates, but meet it well....

[matt.bemis]

Hi @Anonymous,

Thank you very much for the clarification. I am only aware of two different values for torque on an impeller: 1. Via csv file in the scenario folder. 2. Using the Wall Calculator to compute the torque on all the faces of the impeller.

It is important to note that the csv file in the scenario folder is the exact same data set as the "Rotating Region Results" within the GUI. Am I missing a 3rd data set? What am I getting wrong? 

Anyway, I just did a comparison between the Rotating Region results and wall calculator torque, and found very little difference. 

Note that I am not entering a center of rotating; the impeller is exactly located at 0,0,0. All the surfaces of the impeller were previously selected and added to a group for fast selection via Wall Calculator. Considering the results above, I realize there's a difference, but if it's consistently <1% I would say it is non-material. 

During a solve CFD re-centers the rotating region based on its geometry. This is done during the initialization process. I am not aware of the details here but is probably the reason why we see some difference above. I am sure the RR Results uses the newly computed center of rotation.

I am happy to try and get Dev involved here to answer your questions better, but I need more content to provide them 🙂

Thanks,

[Anonymous]

Hi Matt,
If you need some files for testing, get in touch with me.
Well the following is the problem:
"It is important to note that the csv file in the scenario folder is the exact same data set as the "Rotating Region Results" within the GUI. Am I missing a 3rd data set? What am I getting wrong? "
So this is also my understanding...
Normally, there should be no deviation between the csv-file and the results presented in the GUI. But it is there during the run. And this is the problem. I was not sure at the beginning, I thought that the rotating region differs from what is used for the csv-results. Therefore, I checked the pure impeller region vs. the rotating region and this is the 3rd data set (impeller region in GUI , rotating region in GUI vs. rotating region in csv-file). My rotating region has also physically non-rotating wall regions, therefore I made this comparison to see the influence of this non-physical deviation to reality. Then I found out that in every simulation the rotating region results presented in the GUI during run and written in the csv are not the same.
The results which are exported, are coming from the rotating region setting and this should also be in line with the GUI results. But they are not...
If you remember, I asked a few time ago how the variables are computed as I thought of a rounding error or truncation error. This was neglected by Apollo.
I have got now the problem that I cannot trust any simulation, if I cannot see which value is correct. The head calculation (which is correctly calculated) is only a small part for hydrodynamic machine optimizations...
kippo

[matt.bemis · Accepted solution]

Hi @Anonymous,

So I finally have an update. I created a good test model out of the tutorial pump model, and added a lot of mesh and reduced available cores to slow down iterations. This was obviously to allow time to record wall calculator data mid-run. Below is a screenshot of the data table I formed. 

Column 1: iteration #

Column 2: torque value from the Rotating Region Results via GUI. Same as CSV value and obtained at the end of the simulation.

Column 3: Wall Calculator Impeller Torque results recorded during the run. While simulation is running.

Column 4: Wall Calculator Impeller Torque results recorded after the run. The simulation was stopped at 100 iterations.

Column 5: Wall Calculator Rotating Region Torque results recorded during the run. While simulation is running.

Column 6: Wall Calculator Rotating Region Torque results recorded after the run. The simulation was stopped at 100 iterations.

A few comments here:

• There is great agreement between rotating region torque vs. impeller torque while comparing "apples to apples". Example: impeller torque during sim is very close to rotating region torque during sim. Impeller torque after sim is very close to rotating region torque after sim. (great agreement <=1% difference)
• There is massive agreement between Rotating Region Results (csv file) and wall calculator impeller after the run and wall calculator rotating region after the run.
• There is large, unaccepted variance in the torque values via wall calculator while the simulation is running. These may provide guidance, but not hard data for analysis. 

You may next ask me which data to believe and why. I highly suggest using the rotating region data (csv) or wall calculator after the simulation is done. Once the simulation is done, more data is available for analysis and you get a better answer in the post processing. I firmly believe it is not a coincidence that there is almost perfect agreement between the csv file and wall calculator once the simulation is stopped. 

Please let me know your thoughts 🙂 

Thanks,

[Anonymous]

Hi Matt,

this time, I need more time as we will test it live on the test field 😜

O.K. unfortunately, if I do not make the artificial torque correction, then the postponed efficiencies are too low.

I will come back after testing.

kippo