Time Domain Integration velocity to displacement - resulting displacement signal falling

[wlcoyle]

Hello all, 

I am working with an undergraduate student and we are both VERY new to Labview.

We are trying to measure the velocity of a vibrating structure with an LDV and in Labview use the time domain integration in order to get the resulting structural displacement. To test our setup we have a signal generator sending 100 Hz to a simple vibrating beam (clamped free). The velocity waveform we get is reasonable but after the time domain integration the resulting displacement waveform visibly falls off as the measurement continues (please see screenshots). 

A few notes: in the program, please disregard the RMS waveform window. This will not be correct until we get our little problem worked out. Also, we were running this on continuous and this screenshot came after a while of running. When we begin, the displacement (at time 0) is centered at zero and varies sinusoidally but as time goes on, the signal falls and falls. 

Questions:

1) What is creating this fall off?

2) How could we avoid this?

3) Anything wonky you can see in our block diagram?

I have attached some screen shots and Labview block diagrams and front panels that we have for now (simple!) to help in this discussion. 

Thank you!

Sincerely,

Whitney Coyle

[ingram.weeks]

Hi Whitney,

Could this drop-off be related to what appears to be increasing noise present on the velocity sine wave, reducing the area under the curve?

In terms of the block diagram, the STOP Control should be moved inside the structure, otherwise it will not pass a new value to the conditional terminal of the while loop while the test is running.

[Kevin_Price]

This is just like a very common observation when integrating acceleration signals.  Search the forums for accelerometers and integrate and you'll see.

All that's going on is that you appear to have a small DC offset in the velocity signal.  Integrating that produces a negative sloping trend.

Compensation methods vary, here are some to consider:

1. subtract out the average velocity value before integrating

2. do a linear regression after integrating and subtract that line from the displacement

3. If the "DC Offset" drifts a bit rather than being purely constant, you can apply the same idea piece-wise on one cycle at a time.  This is not a mathematically rigorous correction if you are doing any further advanced analysis, but can work pretty well in a seat-of-the-pants way.

-Kevin P