I am looking to record either sound or vibration on an impact wrench (while impacting a tight bolt), and plot the precise delta time between impact events. We commonly use Sig. Acq. for FFTs and many other sound and vibration tests, but this is a new request.
The average impact frequency is approx. 48Hz, but that varies ever-so-slightly from impact-to-impact, and the requestor would like those slight deviations recorded, perhaps plotting precise instantaneous frequency versus impact #, or delta time versus impact # (with impact # being 1-1000 for 20 seconds of data at approx 50Hz).
I really do not want to plot the time-history accel data, and do these calculations manually, between impact peaks.
If anyone has any ideas of functions or calculations to use, I would appreciate suggestions.
Solved! Go to Solution.
I'm not sure I'm understand your questions fully; here are my suggestions:
assume you already acquired the time data? Use signature throughput processing add ins, now you'll see time data processing workbook->Acquisition parameters->change setting
use the setting from above, set a trigger & trigger level, calculations only perform when trigger condition is met; measurement mode is set to track, duration is total length of measurement if you have 100 seconds of data, please enter 100 or greater if you enter 30, Test.Lab will only process up to 30 seconds of data. Increment should be set to the same as the frame size or the impact event delta T, otherwise, it'll include data where there's no impact event.
Go to FS Acquistion->frame size, set the frame size matching with increment
Now you can go to Channel processing->change settings->Vibration or Acoustic tab->Function, you can select FRF, if FRF is selected, reference channel require; or you can select Time, which it'll calculate time block for each impact event base of the trigger settings. Or you can processed the data twice, one with FRF, one with time. Then you can display the processed result on Navigator.
Please let me know if you have more questions.
I initially tried your settings, without saving the waterfall plots, and the Time Data Processing would do nothing. Then, I saved the waterfall plots. This was the result, after collecting 5 sec of data with a 0.5 sec pre-trigger:
I like this. The cursor at 0 sec on the X-Axis shows events at 50Hz, and most of the data falls just below the cursor, since the tool is impacting at approx. 48Hz.
I hope this is what you were referring to for the expected result. I'm sure I could fine tune the settings even further - this was the first iteration.
I have submitted this result to the concept engineer requesting this measurement. If any other questions arise, I will reply back to this forum thread.
Here is an idea for counting the impacts and the time between them.
In the ‘Time Data Selection’ worksheet, display the impacts. Hopefully you can determine a good crossing level to use to define the time between the impacts. In the picture below, using a Single Y cursor, a good crossing level looks like 0.14
Add the function TACHO_PULSE_TO_RPM to the Time Signal calculator. Enter parameters to calculate a 1 pulse per rev rpm with cross level of 0.14.
In Time Data Selection, press “Calculate” (lower left) and press “Save” middle left.
Go to the Navigator worksheet. Drill into the saved data and find a channel in the Throughput that starts with the word “raw”.
Display the “raw” trace. This trace has the number of impacts on the X axis. The time stamp that each impact occurred at is the Y value.
Right click on the trace in the display and choose “Copy Values -> All” and paste into Excel.
In Excel, find the data values (scroll down past the header). The value at 0 is the time of the first crossing (0.623584). Create a formula to subtract the second time stamp from the first.
Pull the formula down. The 3rd column is the time between crossings. Might need to make sure Excel shows enough digits.
Wow. That's a great explanation! Thank you.
Late yesterday into this morning, I have been reading through the Test.Lab help section dealing with Tacho Pulses, specifically the Delta_Pulse function, (realizing it was delta amplitude and not time), in an attempt to make the data easier to read.
I will set that up and report back. Thanks again!
I went through and set up your example using actual test data. On the first attempt, I saw the same issue we have seen in the past while attempting this measurement - a lot of vibrational noise between the impact-socket, socket-bolt, bolt-washer, etc. generating fake peaks/double pulses, and I got this result:
Then, I began looking into the various properties that can be modified while using the TACHO_PULSE_TO_RPM function, and I think the <holdoff_rate> is exactly what I need to clean up the vibrational noise between the actual shock/impact pulses. After calculating the very first holdoff rate percentage, between the first and second pulse, at 50.8%, I then set the holdoff rate to 50%, and here is the result:
Now, that looks pretty good (almost too clean) for just over two seconds of data, measuring a tool with an impact rate of approximately 48Hz.
I'm sure I have to play around a little more with the holdoff rate setting, since I don't think I'm catching all of the impact events. I would bet I'm missing some, since the Throughput data shows impacts above my cut-off of 440g, between 0.76 and 3.13 seconds, so I'm expecting more like 113 pulses, not the 94 shown. Plus, I'm counting 126 pulses in the 'expected' range on the first plot. I'd expect the delta time to be increasing, and the slope to be positive, as the pulse count increases.
You definitely got me started in the right direction! Thank you!
If you have a lot of noise on the data that makes it difficult to count the impacts, you might want to try to high pass filter the impact data before doing the counting. Each impact probably has very high frequency content, while the noise may dominant the lower frequencies. By doing the filter, it may make the impact counting a lot cleaner. You may have to adjust the filter a bit until you get a clean signal in the time domain.
Here is a little big about how to setup a FILTER_HIGHPASS function in Time Signal Calculator
Thank you PJS. I am familiar with the use of filters, specifically Human Vibration filters. I will also look in to using a filter to remove the double-impacts.
I was able to fine-tune the Holdoff Rate, which then returned a number of pulses equal to the delta time x pulse frequency, and the resulting plot looked much more 'real'.
So it looks much closer to the initial graph, with the near-zero data, and has the correct number of pulses I would expect given the time span and frequency of the data.
Now would you have any ideas on why Test.Lab is locking up on me when I save the Time Data Selection, after I use that same function (TACHO_PULSE_TO_RPM) and attempt to use the <Missing Pulse Correction>, to try and fix the few high points in my delta time graph?
I have fine tuned the <Holdoff Rate> to receive the exact number of pulses as peaks in my time-history plot, above the <Cross Level>, and have no problems saving the data, until I activate the <missing pulse correction>. I initially tried a missing pulse correction of 2, and then tried 1, but if it's not set to 0, I cannot save the Time Data Selection.
No big deal, but I thought that correction factor could perfect my data even more.