Bad Data from DCS - Need to Understand 1

[Pavan Kumar]

Hi All,

I am not sure if this the right forum to ask this question. I will ask this question anyway as any help would be great to me.

I trying to analyze data from the historical data from Distributed Control System(DCS) for Make up water flow to Cooling Tower as measured by a flow meter. For many days in the yea r the data is showing as "Bad Data". I wanted to understand what could be the reasons for the DCS to report data as bad data.

Thanks and Regards,
Pavan Kumar

 

[IRstuff]

That's typically under software control, so you, or someone in your company, set up the limits.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[btrueblood]

Sounds like the flow rate was too low to be reliable, a typical problem when flowmeters are run at lower rates than they are calibrated for.

 

[danw2]

The declaration of "bad data" can come from several sources.

If a digital protocol, Foundation Fieldbus or Profibus, is used, then the bus management will declare communication packets as bad data and discard for a variety of reasons, under the umbrella policy that corrupted data cannot be trusted and is always discarded, never salvaged.

If a known device cannot be detected and it is its scheduled time for communication, the result is 'bad device', which might appear as 'bad data'.

If 4-20mA, there could be under or overranges, 3.87mA or 20.12mA that exceed a local limit which is declared bad data.

The field transmitter could be in failsafe mode, at the vendor's declared failsafe value, low at 3.8mA or high at 21.2mA or values close to those.

And I'm sure that there are analog input card/module conditions under which data can be declared 'bad', like an open circuit detection for 4-20mA loop.

Modern instruments have diagnostics which will drive the instrument into failsafe mode under internal component failure conditions. Wiring, connection and electrical noise can cause comm faults.

 

[Pavan Kumar]

Hi danw2 (Industrial),

Thanks for the detailed reply. I came to know today from our electrical department that the flow meter upper limit was too low and that was the reason for the "bad data" report. Now with upper limit increased the issue as told to me has been resolved. However since I am looking at data from 2020, I need help. Now I have another question. The hourly average values calculated for flow rate and shown as bad data - Can it be used?. As you are saying the bad data is discarded, the other data that make up the average in the one hour interval are not discarded with only the out of range value discarded. This means the hourly average data even for the bad data can be used as an approximate number. Is that correct?.

Thanks and Regards,
Pavan Kumar

 

[IRstuff]

Sorry, your post makes no sense; do you have both hourly average and raw data at smaller time intervals? If so, why weren't the hourly averages calculated without the bad data to begin with?

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[danw2]

Some systems use NaN (Not a Number) for bad data which makes it impossible to include a bad data value in a calculation.

All the comm protocols that I'm aware of do not report any useable value when the status is reported as 'bad data' so there's no 'bad data' value to use in a calculation.

However, there is a data acquisition system that reports failsafe high values (21.8mA high or 3.8mA low) as valid points that are pegged 4% of span beyond the upper range limit of device or 4% of span below the lower range value (typically zero for a flow meter).

So there's no global assurance of exactly what is happening unless you have the original data set and can re-run the average.

The answer is specific to your system.

My gut feeling is that your average is likely the average of any available valid data points for the period involved, but it depends on whether your system has a 'valid' data point for 'bad data' status.

If 'bad data' is discarded, then yes, the hourly average data can be used as an approximate number" with the caveat that some dats is missing.