[burhanibrar]

Hello dear Friends, i am still strugglin with this error is CFX if any one solve this please reply. I am doing a transient simulation of wind turbine in which the rotation of blades is due to the forces acting on it because of the wind velocity. I have also tried to edit the RULES file but when i put "ANY" instead of complete line "t, citern, aitern, atstep, " then solver manager does not even start and if i just replace the "t" with ANY then it still gives the same error.

[ghorrocks]

You should not edit the RULES file. Return it to the original version as it is required for the solver to function correctly (as you discovered).

Please explain what you are trying to do and we will help you find the correct way to do it.

Also, I note you posted this same question on other threads. I have deleted those posts. You have correctly started a new thread to ask a new question, so this thread is the place to discuss your issue.

[burhanibrar]

Yes idid as it was before. Acutally i am simulating a wind turbine and the rotation of turbine is due to the forces of the air on the blades. So i have define F which is equal to the resultant force in y and z direction if x is the aixs of rotation then i have use this force to calculate the angular velocit in rad/s and that variable i have defined in the angular velocity of rotating domain. I hope you get the idea and i am very thankful for you concern.

[ghorrocks]

Yes, I understand your question. It is quite a common question and many people attempt to do it as you suggest. It is actually a very difficult and slow way to do it as you now have to get the rigid body motion to converge in addition to the CFD. This is not simple.

That is why the recommended approach in this sort of simulation is to do a range of steady state (ie no rigid body) simulations with a range of rotation speeds and wind velocities of interest to you. You can then define a performance map so you know how much torque your device generates (or absorbs) at any possible flow condition.

You then do a simple ODE solver using your torque map and some system parameters (rotational interia, initial speed, power taken by any devices attached etc) to do a simple model to predict the device speed over time. This ODE model is easily done in matlab, python or even Excel if you have to.

This means the CFD is just a number of simple rotating frame of reference simulations which should be easy to set up, practical to validate/verify and run reasonably quickly.

There are some situations where my recommended approach is not appropriate. If this is the case please explain why.

[burhanibrar]

Thank you again. But i still did not get completely. Actually i have also tried this with immersed solid and in that case it works but when i rotate the fluid domain wwithout solid bodies just the walls of the blade then it wont work. And for steady case the problem is that when i define some velocity let say 10 m/s and what about the rotation of the blades because if i define randomly then the results are different i want to see how fast it rotates at specific wind velocity.

[ghorrocks]

I think you missed the point of my last post. Let's say you want to find the rotational speed your device will have in a 10 m/s wind. You could do it as a rigid body simulation but you will find that to be slow and difficult. My suggestion is to do a series of RFR simulations (ie no rigid body solver) where you assume the rotation speed is 1 rev/s, 2 rev/s and 3 rev/s. If the true rotation speed was 1.5 rev/s then the 1 rev/s simulation will generate positive torque (which shows the rotor will want to accelerate) and the 2 rev/s simulation will show negative torque (which shows the rotor will want to decelerate) and you can interpolate between these results to get the zero net torque result which is the steady state operating condition.

[burhanibrar]

Ok i have understood but it is very time consuming process in the other post you told ssidra that change the rulesfiles or one more option to do this type of simulation. So i just want to ask that the way i choosee to solve it is possible or not. And one last thing that how can i see the torque in Cfx is positive or negative and should i perform trasient simulation or steady one.

[ghorrocks]

You don't seem to understand my suggestion. My suggestion is far quicker than what you are proposing, not slower. My series of simulations can all be scripted up together, maybe using workbench or maybe by command line to run in a single go. They are all simple, reliable easy simulations.

I am not familiar with the details of the RULES file as it is internal to CFX. Users edit it at their peril.

Is it possible? I have heard other users do it, but I have not done it.

Torque direction? Torque is defined based on the coordinate system the torque command was called in.

Transient or steady state? If you are asking this question then you definitely should not do the simulation where you modify the speed. It is not a simulation for beginners. Do the series of simulations I suggested and they will all be steady state simulations, probably with frozen rotor on the GGI.

[burhanibrar]

i have not performed this type of simulation which you are proposing because i have a model of more than 5 millions cell and i think it will take more time because i am doing this in cfx and where i have to do the simulation for each wind speed and for each velocity magnitude. If there is another way please let me know. And i have one more question that in fluent there is option of dynamic mesh 6 DOF and i have seen some videos where the do not define any rotation and the turbine rotates because of the thrust of air does this type of simulation is possible in CFX.

[ghorrocks]

You are overestimating the difficultly of my suggestion and underestimating the difficulty of your method.

The series of simple runs I suggest can all be scripted together and use the previous result as an initial condition for the next one. This means they will run quickly and reliably. And if all you are doing is looking for the operating speed you can probably get it with reasonable accuracy with 3 to 5 simulations. Not very many.

While your suggestion is a single simulation, and is possible, it will be a slow and difficult run as it will require a full transient simulation, will require you to run it until the device gets to steady state and requires the rigid body solver to be converged.

Just because an example shows that it can be done by rigid body solvers does not mean that it is an effective way of doing it.

Also, don't forget the development time. You should be able to start work on the simple way straight away. The rigid body method will require development of the link to rotation speed, and then you are going to have to develop a method to make the rigid body solver numerically stable (just because you link things does not means it is numerically stable and will converge). This is going to take some time to develop.

Before you decide which way you are going to do this simulation, please try a steady state simulation as I suggest, and try a general rigid body simulation. Get a feel for the two approaches. Don't just assume that a single rigid body simulation is simple (because it is not).

[ghorrocks]

Can I make a suggestion?

Try modelling an object falling vertically in free fall. It should end up falling at the object's terminal velocity.

So try modelling this with the rigid body solver - body the body as a rigid body and adjust the inlet velocity based on the motion of the rigid body until it converges on the object's terminal velocity. You don't need any special tricks to do this, you should be able to set this up easily. But it has the same challenges as the rotating model you propose so you will see for yourself what is involved in getting it to work.

[burhanibrar]

Thank you brother but the problem is that i have never done any simulation with rigid body that is why i am confused. I have tried the simulation of free falling sphere today but still did not get the right simulation. If you can help me is rigid body simulation then it would be great like if you have some model then you can send me so i can learn from that. Thank you

[ghorrocks]

That's right, I am not surprised you did not get the right answer. As I have said many times this is not an easy simulation and the approach you are suggesting is very difficult.

I do not have the time to do a tutorial example for you. I have better things to do.

Please try the approach I suggested as it is your only hope of actually getting an answer.

[evcelica]

Just here to confirm that I agree with Glenn. A series of simple steady simulations at different angular velocities will converge much easier and quicker than the complex simulation where the angular velocity is dependent on torque. Simulation like that have always been problematic in my experience.