Redundant Analog Inputs - Really ?

OK Guys,

You have two identical analog inputs that are feedback from a gas pressure regulating system. The objective is to take advantage of both, in case one goes south. The output pressure is, in a word, "stratigic" in nature.

What do you do? I have considered about 10 schemes, and I'm not completely sold on any of them, and the best one is not totally automatic. Anyone have one that is time tested that truely is automatic?

Best Regards,

Bob A.

You haven't provided enough information!!!!

It would be good to know what you have considered. I bet you haven't considered this one. I would use a mathematical model to select which of the two inputs is bad should they differ by more than a set amount.

If the two signals are within some percentage of each other it's easy to just use the average of the two. It's also easy when one shows an open-wire or failed sensor value. In that case, just use the signal from the good one and alert the operator of the failed sensor.

The problem arises when they differ by enough to raise a red flag, but not by enough to make it obvious which one is the bogus signal. If one reads fifty percent of full scale and the other reads thirty percent, is there anything you can use to choose which sensor to believe?

I would suggest 2 out of 3 (2oo3) like in the process world. I would also use different i/O modules.


Nick

Well I see there are some early risers around this morning and that there is a hotbed of activity in the Vancouver area. Thanks for your replies.

Peter, the reason that I did not provide more is because I wanted to see just how "far out" my own thinking was, since I'm an unconventional kind of guy. The fact that you suggested a model tells me that I'm not as far out as I thought. I have used full blown statistical evaluations in the past to determine how to automatically set the gain in a controller, for example. I have considered doing the same to establish which of the two have the best track record. The problem is that it would take some effort to define "Track record" statistically, which involves data collection and analysis, which is not what the customer wants to here in this case.

Nick, thanks for your contributions as I have to admit, neither were amoung those that I had considered exactly. This gas mixing station is at one end of town and the boiler house is at the other about two miles away. In addition to the two inputs at the mixing station, there are several more in the boiler house.

For the first generation, I plan to impliment a combination of the ideas proposed by Steve and Nick. At the mixing station, I'll do the comparison of the two and use the average for the process. When the comparison exceeds a benchmark, I'll annunciate it to the boiler house operator and then he can arrange to override the averaging and select the one that most closely agrees with his own instrumentation.
In phase two, the plan is to incorporate remote I/O from the boiler house to the mixed gas station and then Nick's 2oo3 plan can be implimented.

Thanks again and My Best Regards,

Bob A.

We have pondered this one as well.


RA (and I am sure others) have a SIL-2 compliant module that provides input to multiple input modules.

But in a previous career, we used TMR (Triple Mode Redundancy) and the control system utilized a voting scheme to determine the correct value.

Bob A. said:

Well I see there are some early risers around this morning and that there is a hotbed of activity in the Vancouver area.

Click to expand...

There are two different Vancouvers. Vancouver, WA and Vancouver, BC. Both are named after the same Capt Vancouver. http://en.wikipedia.org/wiki/George_Vancouver
Vancouver, WA is not as well known because it a suburb of Portland, OR.
I have had problems with ignorant US customs people that don't know there is a Vancouver, WA and they thought I was trying to pull a fast one on them.