[lostking18]

Hi, there,

I need to compute the streamline curvature (or equivalently streamline radius) in my 3D turbine domain. Since I didn’t find it directly in the variable list in CFX post and there seems to be very few discussions online, may I know if it is possible to compute streamline curvature in CFX? Either cfx_pre or cfx_post will be good.
One way I can think of is to draw a very dense streamline in the computational domain and export the streamline data. From the streamline data, we can compute the streamline curvature. But it seems to be computational expansive and less accurate. Moreover, after I export the streamline data file, the different threads of the streamline seem to be mixed in the same matrix.

Thank you in advance

[Stel]

I've suggested something similar in a previous thread, but I'll say it again here: in order for you to organize the streamlines you export, together with geometry information select as variables the Streamline Number on Streamline 1. Save the file as a .csv. Open this csv file in Excel, and sort all the Data in ascending order using as a reference the column where the Streamline Number on Streamline 1 values are located. This will separate individual streamlines. Maybe this can serve you any purpose.

And why do you think that calculating curvature with exported streamlines will not be accurate?

[lostking18]

Dear Stel:

Thank you very much for your reply. The suggestion on Streamline Number is really helpful!
Since my curvature is 3D, I need to calculate the derivative of the curve to get the curvature. This leads to the discretization error. Do you have better options to compute the curvature? Thank you in advance.

[Stel]

Two suggestions: 1 - perform a rigorous mesh sensitivity test to investigate how the local velocity gradients (and eventually the curvature of the streamlines calculated with CFD Post) are sensitive to mesh refinement; 2 - use the streamline definition to create a field variable which can be used to calculate the curvatures as a function of local quantities, and not the CFD Post streamline feature. That will require you to revise some definitions about how a streamline is defined. If this proves successful, you will have the streamline curvature defined in the whole flow field, and not along individual streams, which can be very useful for you.

Good luck!

[lostking18]

Originally Posted by Stel

Two suggestions: 1 - perform a rigorous mesh sensitivity test to investigate how the local velocity gradients (and eventually the curvature of the streamlines calculated with CFD Post) are sensitive to mesh refinement; 2 - use the streamline definition to create a field variable which can be used to calculate the curvatures as a function of local quantities, and not the CFD Post streamline feature. That will require you to revise some definitions about how a streamline is defined. If this proves successful, you will have the streamline curvature defined in the whole flow field, and not along individual streams, which can be very useful for you.

Good luck!

Dear Stel

Thank you for your answer. Your second suggestion is intriging. But I'm not sure what exactly you mean. May I ask it more specifically?

What do you mean by 'use the streamline definition to create a field variable which can be used to calculate the curvatures as a function of local quantities'?
Do you mean we can create a new variable in CFX Post to give a new definition of the streamline?
Or do you mean we can create a variable in CFX Post to compute streamline curvature regardless of the definition of the streamline?

Why do we need to revise the definitions of the streamline? How can we revise it as I didn't see this option in CFX Post. Do you mean we need to revise it in CFX Pre?

Sorry for the long question. I tried to understand it but didn't grasp it well.

Thanks!

[Stel]

There is a way to calculate streamlines not using time integration, but actually with pathlines (paths that are tangent to the velocity vector at each point). I did this once in the PHOENICS software, but unfortunately that was a long time ago and I don't remember it well to give you detailed advice (but I know that it is possible). What I was telling you is to revise fluid mechanics materials to recall this definition in order to think of another way to calculate what you want. That is, maybe you are stuck into this paradigm that you actually need to draw a streamline to calculate what you want, but maybe it is possible to derive an algebraic expression to obtain a curvature of a local flow path using local variables.

But forget all this if it seems too complicated and confusing and go for suggestion 1, which I believe you can do with what you have in your hand right now (something is better than nothing...).