Are PCI DAQ cards suitable for 24x7 use ?

[MogaRaghu]

I have an application that needs to run for 6 to 8 months non stop .

Its an Endurance rig for cycling a solenoid valve and plotting the switch ON / OFF times. The main code ( LV 2009 ) runs inside a 10mS hardware timed loop.  With a switching frequency of about 1 Hz,  I need to cycle for about 20 million times. (There are no missing Iterations as the average time taken is about 4-5ms only.)

Using PCI 6229 and PCI 6514 cards inside an Industrial PC from Advantech and running on WIN-7 platform.  

So far the system has run for about 3 weeks without issues. Will this run for rest of the period ??  In other words is my approach correct with the mentioned combination of H/W and LV2009 ? Or whether a RTOS with cRIO be more suited here ??

Thanks

[JÞB · Accepted solution]

Running long term test can be very challenging!  There's just too many things that can go wrong- Power failures, equipment malfunction, operator errors and even managers dropping a cup of coffee in the set-up (trust me, its happened).  

When I run across this type of requirement I try to think of how to "fail-safe" the system so that, even if the PC explodes I can recover the data and continue the experiment.  In your case it does not sound as if EVERY one of 20M cycles is absolutly needed.  You could still learn about the solenoids life cycle from the trends in the data.  I assume you are interested in break-in, long term trends, and EOL indicators.  So logging data regularrly with an additional event log (event: system halted - Power faulure 10,787,690 cycles complete at YY/MM/DD/HH:MM:SS etc.) would probably keep any conclusion you draw from data analisys valid.

But most of that is software requirement design and design for fault tolerance.  The hardware you list sounds suitable for this experiment (And pretty close, as I can tell from the limited spec discription, to what I would have looked into for this type of solution.) 

[MogaRaghu]

Hi Jeff,

That was pretty reassuiring to say the least.

You are right in assuming that missing a few cycles in between is not going to matter much as I am interested in the long term trends.

>>So logging data regularrly with an additional event log (event: system halted - Power faulure 10,787,690 cycles complete at YY/MM/DD/HH:MM:SS etc.)

This will call for some kind of supervisory system over and above the PC running my code ?? Right now I do a 10 cycle profile capture and update the cycle count to disk once every 5 minutes so the worst case loss is just about this amount. Of course is the disk crashes then the last back up on removable media is the saviour. Maybe a secondary hard disk could save the day.

[JÞB]

Yes! for long term test you need a supervisory loop WITHIN your solution! to monitor-

• A UPS with connection to the PC (shut-down on power loss)
• Network events (Sever errors-  you need to backup the data!)
• user interupts (think Planned IT upgrades or calibration)

Glad to have helped!