Hardware setup:
We are putting our first NI9237 module into the test lab, to replace the old SC-2043-SG board. For now, the two PC's are running side-by-side using the same load cell. The old PC is running Windows XP and a load-acquisition program written in LabView 7.1 and using the SC-2043-SG. The new PC is running Windows 7 and a similar load-acquisition program written in LabView 2012 and using the module NI9237. Both PC's are cabled to the same full-bridge load cell (one at a time - not in parallel)
The problem:
When we test our parts, we apply a sudden increase in load, using our pnuematic tool, until the part fractures. The output characteristic is the peak load which we call 'breakload'. The two PCs are giving statistically significantly different results. The two PC's are showing a 7% difference in peak load capture when used in our dynamic testing with actual test parts. (e.g.: old PC 1808 lbs average, versus 1942 lbs average for the new PC. We have tested many parts - far too many for this to be natural variation.)
What we need from National Instruments:
We need to know why the two modules are giving different results. It's not so much that we need them to agree - but we need to know that the results are valid, coming from the new PC/LabView 2012/NI9237 system. If they are different, we need to know why.
Would you please look are our source codes and see if there is any reason for this difference. Also, please let us know if there are any differences between the SC-2043-SG and the NI9237 that might cause this.
What we know so far:
Our calibration technician has verified that both modules are in calibration - using our same load cell and comparing to his master cell, but using the two different modules (SC-2043-SG & NI9237) for each calibration exercise. The calibration exercise is a series of static applications of loads to the subject load cell in series with the master cell.
But the two PC's are showing a 7% difference in peak load capture when used in our dynamic testing with actual test parts.
Also, the calibration technician has compared the peak-capture algoritms of the two programs by "bumping" the load cells with his hydraulic cylinder. Both the old and the new PCs agree with his peak-hold reading on his master cell. However, when he 'bumps' the load cell his load is applied over about 1/2 to 1 full second. When we acquire our loads in our test lab. the full duration of the application rarely lasts longer than 150 milli seconds. (We are sampling at 1000 per second.)
Attachments:
"graph.vi" is from the old program in LabView 7.1 and earlier.
"CaptureGraph.vi" is from the new program
"W53-20130531-0002.graph" is a text file that contains the data captured during a graph acquisition. In this case the breakload was 1819.177 lbs. The raw data is listed after the "">>>>" marker. It was acquired at 1000 samples per second.
Contact info:
Please RSVP to me (John Moodey) and copy the following emails. I appreciate it very much:
<Emails and contact information removed>
Thank you in advance for your help in this very important issue.
John Moodey, Quality Engineer, Cherry Aerospace
