Reference and measured path comparison

[simasa]

Hello,
I am planned to calculate the deviations in reference and measures tool path of a milling machine (for offline analysis in computer).

Reference path is generated using reference CNC code and measured path is measured using sensors on TCP (tool centre point)

Both paths may looks like this:

Reference = [time x y z] points up to 20,000
Measured = [time x y z] points up to 50,000

I guess I can not simply calculate the deviations taking time as reference:
Deviations = Reference - Measured (at same clock time)

Because:
Machine controller (simumeric 840D in my case) may introduce small uneven/variable time delays over the whole path making which finally appear in measured path (making time based comparison not reliable)

Any idea please? How such problem is addressed typically?

Regards
simasa

 

[IRstuff]

I don't understand what it is that you're asking. What type of "deviations" deviations are you looking for?

Seems to me that you'd interpolate into the reference dataset and find the least squares deviation.

TTFN

FAQ731-376

 

[simasa]

hello
Sorry I did not explain the deviations.
A milling machine has a certain accuracy lever (finishing tolerance). This is due to the dynamic and static errors in a milling machine. More the dynamic and static errors are in a milling machine less is the accuracy. In other words more would be the position deviations in TCP (tool centre point).
Thus if we measure the TCP (measured path) and subtract it from the Reference path (CNC code), we get the deviations.
These deviations can be a measure of the machine accuracy.
I record both reference and measured paths of a machine and now I want to calculate the deviations by subtracting both paths.

But I do not know the way to find the corresponding points in both trajectories, as I think there would be time shifts and delays in measured trajectory introduced by the controller.

It might be a typical problem. Thus I hope I could get some expert suggestion.

Regards
simasa

 

[IRstuff]

You're still not making sense. You claim that both sets of data are xyz, so where does time come into play?

You should be doing a correlation of sorts to find the best fit offset. Once you find the offset, you determine the degree of fit.

TTFN

FAQ731-376