open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I have just noticed a conflit in the way Bluekenue treat VELOCITY_DIRECTION.
In one hand, the direction corresponds to the second column of the velocity time series at one point and in the second hand the direction can be extracted from the calculator (A=VELOCITY_DIRECTION) and then makes the object of a time series at the desired point.

I made sure (several times) that the time series were extracted at the same point but I still see a discrepancy between the two modes of extracting velocity direction time series.

Someone could explain me where this comes from and which one I should believe?

Thanks in advance.

With best regards,


Alexis

Hi Alexis,

I just checked into what I think you are describing and I am getting exactly the same direction using both methods.

Just so that I'm clear on what you are doing
I'll describe what I think is your procedure.

First method
1. Select a node of a VITESSE UV object. (eg. Node 100)
2. Extract a time-series. This gives you a ts2 time-series containing magnitude and direction.

Second method
1. Select the VITESSE UV object in the workspace and open the calculator.
2. Set Variable A=VITESSE UV.
3. Select Dir. from the drop-down.
4. In the expression field type A and set the Result Name to "MyDir"
5. Press evaluate. This results in a new time-varying scalar t3s file.
6. Select the MyDir object in the workspace.
7. select the Node 100 and extract a time-series giving you a ts1 time-series containing direction only.

Are saying that the dir value in the first time-series is NOT the same as the dir value in the second?
I just walked through these steps and I see no discrepancy...

Can you post you selafin file and tell me which node you are selecting? I'll try to duplicate the behaviour you are seeing with your data.

Cheers... Martin

Martin,

Indeed the time series are the same when you extract them at a NODE but they differ when you take them at a POINT via Tools->Extract TimeSeries At Multiple Points ...(an interpolation is necessary).

How can this happens?

Thanks a lot!

Alexis


PS: It does the same for the VELOCITY_MAG time series.

Hi Alexis,

When you use the "Tools->Extract TimeSeries At Multiple Points" function the values will be interpolated (linearly on the element) at the location you specify in the supplied pointset.

If the coordinates in the pointset are not exactly at the same location as a node, I would EXPECT to see some differences. These differences will increase, the further away you are from the node. Is this perhaps what you are seeing?

What is the relative size of the differences?

Cheers... Martin

Martin,

There is also an interpolation when you extract a time series at a point with method n°1. The thing is that these two interpolations are for some reason (???) different.
For instance (at A specific point of my estuary where I have also instruments) the velocity direction time series (with method 1 and 2) vary slighty in average (less than 1 deg) but their discrepancy can reach 48 deg sometimes !

Which of the two methods do you think is better ?

Thank you!

With best regards,

Alexis

Hi Alexis,

If you extract a time-series from a selected node (method 1) the data comes directly from the Selafin file. No interpolation is performed.

A discrepancy of 48 degrees is certainly unexpected. I think I'm missing something in your description. Perhaps we should take this off-line and communicate via telephone or email to resolve this. Debugging this through the forum is too slow. Once we've resolved the problem we can post a final resolution on the forum for the benefit of others.

Cheers... Martin

Hi Alexis,

Aha. Now I understand what's going on...

It comes down to the fact that direction is not linear on an element.

When you extract a velocity time-series, at an arbitrary point "P", from a "VELOCITY UV" object, Blue Kenue does a linear interpolation of U and then V in the element and then calculates the magnitude and the direction.

If on the other hand you use the calculator to generate a scalar surface of "directions" and then extract a time-series at the point "P", you are telling Blue Kenue to linearly interpolate "direction" on the element.

These 2 calculations are not equivalent.

So, as to your question "which method is more accurate?", what you really want to ask is "which method is appropriate", I say you probably will want to use the time-series extracted directly from the VELOCITY UV object.

In the future what I suggest is, when you are building the mesh you should include a hard point at the location of your field instruments.

If I'm not explaining this properly perhaps Jean-Michel can add a comment...

Cheers... Martin