Signal Trending and Data Historian

You're not wrong, but as an engineer, you design to requirements and desirements from stakeholders. Unless a stakeholder actually writes down a requirement, you have no basis to add capability; now, you can try and convince a stakeholder that historical data is useful. I don't understand why your process engineers don't already see that utility; seems dubious to try and do process control and improvement without data.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! faq731-376 forum1529 Entire Forum list

Brown Instruments (subsequently bought by Honeywell) ran this ad in the late 1930's or 1940's for its then-high-tech strip chart recorders, the original 'historians'. Without a history, how do you account for the gremlins?

Lots of times I've seen idiotic data capture and storage that could've only been there because some fool listened to some ignoramus who said, "You should record this forever so you can look back and solve all your problems."

You can't rationally use the same sensors your process is using because when they start drifting and your process is going to hell you have no clue because it all 'reads correctly'. This then requires an entire second sensor suite that then requires another set of calibrations to be maintained. Then you need a way to record gobs of data FOREVER. Data that may never ever be looked at.

One needs to be very thoughtful about whether or not they really need to burden the system with data storage. I'd guess about 98% of the time it's completely uncalled for.

Keith Cress
kcress -

Wow, I finally disagree with Keith; a first.

I can't tell you how many times on the PLC forums an OP is asked, "can you provide a trend so we can see what's happening?" and they can't. A graphic trend fulfills Confuscious' "Picture worth 1,000 words" saying.

As to drift, the only common sensors that really drift are thermocouples, the higher the temperature, the more the drift. The industries relying on those for accuracy (aerospace/military/automotive) are spec'ing exposure incidents and temperature limits; making base metals T/C's much more disposable than in the past. And some people don't know thermocouples drift and use a J at 1600F for 13 years and wonder why they have problems.

RTD's are stable unless they're in high vibration that short a wire wound element.

Pressure transmitters are incredibly stable.

The electronics on modern flowmeters are generally pretty stable, it's the mechanics that take a hit: the sharp edge on an orifice plate erodes over time, but the pressure transmitter holds up with negligible drift, the bearings on turbine meters need periodic replacement, electrodes on mag meters can corrode or coat depending on the medium. One thing I liked about Siemens clamp-on ultrasonics was that if it couldn't give a reliable reading, it gave a fault reading, which is not uncommon when used on old pipe with an unknown internal surface condition (but they didn't drift). Coriolis doesn't drift but it does not auto-zero, but that's no secret, it just has to be taken into account in the process. Running DP flow outside the design conditions for gases in particular, without pressure and temp compensation is asking for incorrect measurement, but that's no secret either. The means for compensation is there, it just needs to be used.

The diceyness of continuous level measurement is evident by the rule to have a different technology used for a high limit, point level switch, to make sure the continuous level instrument isn't reading something other than reality. So, yes, continous level is dicey, but who records level? Not many people because who cares what the level was 3 weeks ago? Or even yesterday? The purpose for level indication is know whether there's enough stuff in the tank/vessel to make the next batch or whether someone's gotta call the vendor for another delivery.

Load cells can and do drift, except for those zero-drift German etched ones, used to be Sartorius, now some other name.

Not that every process signal in the world needs to be archived in an historian, but there sure are lots of times I've dealt with situations where what has been happening is all anecdotal because there is not history.

Like video security tapes, process data has a limited life; after some time, it's no longer of interest, PROVIDING, the trend data is kept.

I don't thing Keith is advocating keeping zero data; he's advocating a targeted data and intelligent data collection process.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! faq731-376 forum1529 Entire Forum list

Thanks IR.

Yes, that's why I said 98%.

There are certainly places it's needed. There are some processes that require it period. For instance the sterilizer industry has to log everything or product can't be considered processed without that historical record. There they either need separate sensors or they have to have a nuanced instrumentation system where calibrations are done frequently and meticulous records are kept so a 'drift' type event can be tracked back and applied to the record if necessary.


You're right Dan about T/Cs but then they're ubiquitous in industry so that drifting is available frequently.

I'm just saying if you log that same sensor you historical record is garbage at best. Your product slowly gets a gooey core that grows food poisoning but your record sits there telling you the temperature is just fine like your process controller thought it was. That all actually makes the entire process of finding and correcting the issue harder because you have the facts wrong.

Dan, can you give a few examples of where a data historian would actually be useful?




Keith Cress
kcress -