Learning Forum

[Viktor Merkel]

Hello everybody!

I have made a CFD simulation of an axial turbine with CFX and I have evaluated the power generated by the turbine by two different approaches:

1. Torque based: This means that I have used the built in function (of CFD-Post) torque_z@Airfoil to evaluate the torque generated by the airfoil (while z is the rotational axis). I did the same for the hub surface, summed them up and mulitplied with the angular velocity. This gives the generated power for one stage. I do the same for all other stages and sum up all values which gives the generated power for the whole turbine.
2. Total Enthalpy based: I do straightforward balance of the total enthalpy. This means that the difference of the sum of all icoming total enthalpy streams and the sum of all outgoing total ethalpy streams gives the generated power. If I do the balance for stage, I get the stage power. If I do the balance for the whole turbine, then I get the power for the whole turbine. 
   For the evaluation of the incoming and outgoing total enthalpy streams, I use mass flow average of total enthalpy at the boundaries of interest.

My point is that the power evaluated by the two approaches deviates by 2% to 3% (while torque based is smaller for all evaluated cases).

I would like to understand the following points:

• Is this a usual range? Or do I make a mistake somewhere?  
• Which of the two approaches is more trustworthy?
• What happens with difference if I refine the mesh significantly? Will go to zero?
• What is the best practice?

Thanks for any answer!

General comment: If someody thinks that the deviation between the two approaches is too small to be concerned about it, I would like to reply that it makes about 2...3% in efficiency, also! Taking the higher value means that I have already reached my goal. Taking the lower value means that I still have to do a lot of improvement on the turbine.
 

[CFD_Friend]

Hi Viktor,

The torque method relies on integrating forces. Small errors in force distribution, integration limits, or near-wall resolution can lead to underprediction.
The total enthalpy (energy balance) approach, on the other hand, is based on a global conservation principle. It is often less sensitive to local grid resolution errors.

Refining the mesh typically reduces discretization errors. With a finer grid, you can capture gradients (especially near the blade surfaces) more accurately, and the two methods’ results will likely converge closer.

[Viktor Merkel]

 

Hello CFD_Friend,

thanks for comment. I totally agree! The both methods should give the practically the “same” value.

I also understood why I had such big differences. As I was simulating the expander of a gas turbine, I split the airfoil surface at the trailing edge and set an inlet there instead of a wall. This inlet accounts for the cooling air that is coming out of the trailing edge / cutback of a cooled gas turbine blade. 

The point is that the torque evaluated at such an inlet is probably not very accurate. I made a test run at which I switched the inlet to a wall-boundary and the deviation between torque-based and tot.-enthalpy-based power decreased to below 0.1% which is absolutely accutate enough for me!

My conclusion is that one can use torque-based method if one has no inlets at the airfoil surface and as it less effort, I will do this per default. But if one has a more complicated case (with inlets or any other stuff), one should use tot.-enthalpy-based method as it is more accurate!