CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   STAR-CCM+ (https://www.cfd-online.com/Forums/star-ccm/)
-   -   Time-step for Implicit Unsteady solver in STARCCM+ and maximum number of steps? (https://www.cfd-online.com/Forums/star-ccm/226348-time-step-implicit-unsteady-solver-starccm-maximum-number-steps.html)

mazhar16823 April 24, 2020 07:50

Time-step for Implicit Unsteady solver in STARCCM+ and maximum number of steps?
 
Hi,


I am working on the CFD simulation for a wind turbine blade. I chose Implicit Unsteady Solver for this purpose. I want to associate the time step with rotational speed of the blade I am unable to figure out how should I do this and what should be the max. physical time and maximum steps for the stopping criteria? Any comments please.

SimulatorMaster April 25, 2020 08:38

Hey there,

I have done a case where I had a spring valve closing and opening. The way I determined time step and inner iters. were somewhat trial and error but my procedure was like this (Im sure there are some kind of theoretical calculations involving this matter but I do not have the knowledge):

Initially, calculate how far ur blade travels for each time step and according to your mesh, either coarse or fine, choose your time step.

For example, lets say your mesh is really fine and has a size of 1mm and in 0.001 second it rotates by a value greater than 1mm, say 10mm (calculations should ofcourse be in radians and all that just an example). You should reduce your time-step in this case to be close to 1mm. (The deformation of mesh should not be a huge, sudden jump) As for the inner iteration, start with 7, and if your residuals drop below 10^-3 or lower, you can leave it like that or you can increase inner iters if you would like.

This worked pretty good for me but ofcourse this is an entirely different case.

Hope this helps. =D

mazhar16823 April 25, 2020 08:57

Quote:

Originally Posted by SimulatorMaster (Post 767327)
Hey there,

I have done a case where I had a spring valve closing and opening. The way I determined time step and inner iters. were somewhat trial and error but my procedure was like this (Im sure there are some kind of theoretical calculations involving this matter but I do not have the knowledge):

Initially, calculate how far ur blade travels for each time step and according to your mesh, either coarse or fine, choose your time step.

For example, lets say your mesh is really fine and has a size of 1mm and in 0.001 second it rotates by a value greater than 1mm, say 10mm (calculations should ofcourse be in radians and all that just an example). You should reduce your time-step in this case to be close to 1mm. (The deformation of mesh should not be a huge, sudden jump) As for the inner iteration, start with 7, and if your residuals drop below 10^-3 or lower, you can leave it like that or you can increase inner iters if you would like.

This worked pretty good for me but ofcourse this is an entirely different case.

Hope this helps. =D

Thanks for your input. Your case also seems reasonable. But I don't understand you said ''calculate how far ur blade travels for each time step and according to your mesh'' because I can only see how a parameter for instance velocity behaves across the blade because I have given one input that is incoming wind velocity and the rotational speed of the blade. How would you examine if it rotates a value greater than a certain number? My mesh is 0.005 m on the blade.

SimulatorMaster April 25, 2020 10:04

I assume that your turbine(your solid region) is in another region that is a fluid rotating(perhaps you used overset mesh?). What i mean by that line is that your turbine will rotate at a specific amount according to your time-step. For a specific time step, x, your blade will rotate a specific amount, y. So if it turns at 1rad/s, for a 0.001second time step, with a radius of 2 meters lets say, the tip of your blade will travel 0.002m or 2mms. 2mm travel is pretty acceptable if your mesh size is around 1mm or so AT YOUR BACKGROUND MESH.

The mesh on your blade is important yes, but whats really important in this case is the mesh on your background region, which is your fluid region.

So, what you need to do initially is to calculate the maximum rotational speed that your turbine will achieve with a given fluid velocity. Then, calculate your maximum displacement at max. rotation then proceed with the calculation I did above. Finally, find your timestep.

You can also calculate the rotational acceleration of the turbine and find your maximum physical time needed to achieve max rotational speed. (There should be equations lying around=P)

Furthermore, if you are going to analyze the deformation on your wing, (just throwing it out there), after you setup your morphing(a motion option), mapped interfaces and all that junk, adjust your mesh similar to your background mesh (your fluid region). The motion Tutorials on StarCCM really help.

Hope I was clear. Though if you have more questions please ask. =D

mazhar16823 April 25, 2020 10:46

Quote:

Originally Posted by SimulatorMaster (Post 767331)
I assume that your turbine(your solid region) is in another region that is a fluid rotating(perhaps you used overset mesh?). What i mean by that line is that your turbine will rotate at a specific amount according to your time-step. For a specific time step, x, your blade will rotate a specific amount, y. So if it turns at 1rad/s, for a 0.001second time step, with a radius of 2 meters lets say, the tip of your blade will travel 0.002m or 2mms. 2mm travel is pretty acceptable if your mesh size is around 1mm or so AT YOUR BACKGROUND MESH.

The mesh on your blade is important yes, but whats really important in this case is the mesh on your background region, which is your fluid region.

So, what you need to do initially is to calculate the maximum rotational speed that your turbine will achieve with a given fluid velocity. Then, calculate your maximum displacement at max. rotation then proceed with the calculation I did above. Finally, find your timestep.

You can also calculate the rotational acceleration of the turbine and find your maximum physical time needed to achieve max rotational speed. (There should be equations lying around=P)

Furthermore, if you are going to analyze the deformation on your wing, (just throwing it out there), after you setup your morphing(a motion option), mapped interfaces and all that junk, adjust your mesh similar to your background mesh (your fluid region). The motion Tutorials on StarCCM really help.

Hope I was clear. Though if you have more questions please ask. =D

Firstly, I only have the fluid (static region) around the blade. The domain is 3D in bird view (V-shaped) where blade is placed on extruded-hub. I am not using overset mesh. My purpose it to capture flow transition not the deformation. Also, I already have the rotational speed for the blade and the incoming flow velocity (which is obviously needed to create the report).

This is still confusing '' For a specific time step, x, your blade will rotate a specific amount, y. So if it turns at 1rad/s, for a 0.001second time step, with a radius of 2 meters lets say, the tip of your blade will travel 0.002m or 2mms''. As I have already provided the rotational speed in reference frame how will I know that at what rate the blade is turning in a specific time step?

SimulatorMaster April 25, 2020 16:29

Seems like I made unnecessary comments based on your domain. If I understand correctly, your rotation is constant (no acceleration) which means its rotation rate will be constant for each time step, say 1rad/s.

If you have rotational acceleration, you can create a report in reports tab and create a motion report and choose your function accordingly. Then monitor and plot this report to see what your rotation speed is at each physical time. Otherwise, for a constant acceleration, it wouldn't make sense to monitor the rotation rate. (Maybe displacement)

=D If you are still confused, just give it a time-step and let it run and keep an eye on your residuals. Reduce it if they go crazy.

mazhar16823 April 25, 2020 16:41

Quote:

Originally Posted by SimulatorMaster (Post 767360)
Seems like I made unnecessary comments based on your domain. If I understand correctly, your rotation is constant (no acceleration) which means its rotation rate will be constant for each time step, say 1rad/s.

If you have rotational acceleration, you can create a report in reports tab and create a motion report and choose your function accordingly. Then monitor and plot this report to see what your rotation speed is at each physical time. Otherwise, for a constant acceleration, it wouldn't make sense to monitor the rotation rate. (Maybe displacement)

=D If you are still confused, just give it a time-step and let it run and keep an eye on your residuals. Reduce it if they go crazy.

You're right about rotation rate it's constant.But for my study I want to give several time steps and see on which time step I get steadiness of the results.
For instance, I was assuming to make calculations by considering the specific rotation degree in one full rotation of turbine blade i.e. 360. For example, if I take 2 degree rotation and say 4 number of rotations and then using this information I calculate time-step size and number of time-steps. It would be a good idea but I am just concerned about the choice of ''degree rotation'' and ''number of rotations'' that on what basis I should choose them - I mean there must a reason for choosing a specific degree rotation and number of rotations. Would appreciate your feedback on this.

SimulatorMaster April 26, 2020 09:44

Yea well said friend. I have dealt with overset mesh which was a pain in the a** but I managed somehow. You do need some experience through trial and error and, ofcourse, man hours.

Here is theh thing about time-step; I choose it accordingly after few trials while monitoring residuals. Again, you do not want a huge time step for better accurate results. Like my friend said "...the movement of the blade should not exceed 0.5 cells."

You should not base your rotation rate on your time-step; should be the other way around. So, choose a rotation rate, and then choose your time step according to whatever mesh you have. If you would like to implement acceleration, find the max displacement over a second, and again choose your time-step.

If you would like to analyze the change of increasing time-step on your domain, you can assign functions that will automatically do it for you. (Or just do it by hand after a full rotation.


Since your flow is constntly flowing but your blade is rotating, you wont really reach a steady state. What important are you residuals.

Assign a time-step and give your sim a go. First step is always important. =D

Im sorry if Im confusing. Im not an expert Im just giving out my knowledge from my experince through overset mesh and motion sims.

mazhar16823 April 26, 2020 10:23

Quote:

Originally Posted by Steklo_Plastik (Post 767379)
Your path is as old as my life. Look for another way. It should be with natural accompaniment. Take DFBI and 6 dof. That is happiness. Otherwise, the wind will not loosen your screw. You will have to set the fan speed. This is a bad and false way.
Specifically for your question: the movement of the blade should not exceed 0.5 cells.
You can use the automatic control model CFL. Download from physical models. If the accuracy of the calculation is important to you, and not the beauty of the animation, then use it. Draft calculations can be done in large increments, but no more, when the calculation is destroyed. It is established experimentally.
Good luck and fast growth.
I warn you: the program is not yet perfect in terms of 0versetMesh . It is difficult when there are many bodies.

https://youtu.be/PigTaG_GFoo
https://youtu.be/4fk2vH2pm-E
https://youtu.be/587oz1d41KU

Auxiliary images:
Attachment 76955

Attachment 76956

Attachment 76957

Attachment 76958

Attachment 76959

Problems only occur when the rotating 6 DOF bodies are connected to the main body. We need to gain experience. There are no lessons here. Insert 6 DOF - (DFBI) in scenes, not in surface meshes of bodies.


Thanks. But my case is related to onshore WT blade not offshore so I have nothing to deal with drafts etc.

mazhar16823 April 26, 2020 10:54

Quote:

Originally Posted by SimulatorMaster (Post 767418)
Yea well said friend. I have dealt with overset mesh which was a pain in the a** but I managed somehow. You do need some experience through trial and error and, ofcourse, man hours.

Here is theh thing about time-step; I choose it accordingly after few trials while monitoring residuals. Again, you do not want a huge time step for better accurate results. Like my friend said "...the movement of the blade should not exceed 0.5 cells."

You should not base your rotation rate on your time-step; should be the other way around. So, choose a rotation rate, and then choose your time step according to whatever mesh you have. If you would like to implement acceleration, find the max displacement over a second, and again choose your time-step.

If you would like to analyze the change of increasing time-step on your domain, you can assign functions that will automatically do it for you. (Or just do it by hand after a full rotation.


Since your flow is constntly flowing but your blade is rotating, you wont really reach a steady state. What important are you residuals.

Assign a time-step and give your sim a go. First step is always important. =D

Im sorry if Im confusing. Im not an expert Im just giving out my knowledge from my experince through overset mesh and motion sims.

Thanks for the comments. I did the first run but I see that the residuals started showing up from the right side of the plot rather left side which is uncommon for me. Any idea what could go wrong?

SimulatorMaster May 10, 2020 11:27

Huh thats wierd. Sorry I didnt see this were you able to overcome the issues? A reason for that would be the plot settings on residuals graph which can be editted by right clicking on the axis.

mazhar16823 May 10, 2020 11:29

Quote:

Originally Posted by SimulatorMaster (Post 769702)
Huh thats wierd. Sorry I didnt see this were you able to overcome the issues? A reason for that would be the plot settings on residuals graph which can be editted by right clicking on the axis.

Thanks. I had resolved this issue. :)


All times are GMT -4. The time now is 12:04.