4 replies to this topic

[thorium90]

Hi all,

 

This question is regarding the accuracy of flowmeters, more specifically electromagnetic flowmeters, although the concept can be generally applied for other similarly installed plants.

 

Say there are multiple parallel trains in a plant and there is a flowmeter measuring the output from each train in this plant. All of these outputs go to a common header which has another flowmeter of the same type that measures the combined flow.

Assuming all other factors are the same, fluid properties, instrument specifications, etc, which would yield a more accurate measurement? The sum of all the flowmeters which would give the total flow, or the single (larger, since it is in the combined header)  flowmeter?

I have heard differing view points, thus far on which method gives the more accurate measurement. So, therefore I'm putting it out to see what are the thoughts of others.

[Pilesar]

Find the expected error of the flow instruments. Accuracy is rated as a 'percent of calibrated span'. Smaller flows are often measured by instruments with smaller spans. If the instruments are the same type with the same 'percent of span' error, then the smaller flows measurements should be more precise and the sum of the smaller flows would be more precise than the single larger flow measurement. The most accurate flow indication is to use all the available meters and reconcile the data using standard data reconciliation principles taking into account the expected error of each instrument. Redundant instruments can result in significantly better accuracy. 

[thorium90]

Hmm, the reasoning I got for why parallel flowmeters would have more accuracy was that that if the flowmeters are assumed to have a measurement uncertainty of normal distribution, then the bias associated with each flowmeter would not always be in the same direction or magnitude and hence parallel flowmeters would exhibit lower uncertainty in measurement than a single flowmeter.

What do you guys think? Is the statistical reasoning valid? The reasoning seems sound in my opinion.

[Bobby Strain]

Maybe. But not necessarily true of a particular installation. But it is certain that parallel meters are less reliable than a single meter. One would likely choose the most accurate single meter if this is a basis for transactions. Lesser accuracy for each train would suggest less expensive meters for each train.

 

Bobby

[Saml]

"then the bias associated with each flowmeter would not always be in the same direction or magnitude and hence parallel flowmeters would exhibit lower uncertainty in measurement than a single flowmeter"

 

You are right, but not totally. It is right that the probability of deviations in the same direction is low. But why discard the total measurement?. You can do data reconciliation and use all the data you have available. It also have the advantage of detecting a gross error in one reading.

 

In the site of the Montreal Polythecnique there is this course with good information. It is the source I've found more friendly to read.

http://www.polymtl.c...iationTech.html

 

There is also a book "Data Reconciliation" by Sanchez and Romagnoli, that is very good. However I have found it a little bit harder to read if you are new to the topic.

 

Excel lacks some matrix algebra support you may need, but free tools like GNU Octave can do the job efficiently. If you have Matlab is even better.

 

The other way is to go the brute force way with Excel. You set a column of readings, a colum of standar deviations a column of reconciled values and a last column of cuadratic difference diveded by each stardard deviation . Then you use a solver to minimize the sum of these differences by changing the reconciled values,  subject to the condition that the total is the sum of the partials. I this case, it may be possible to go this way.

 

In every case the key point is defining the accuracy of each meter.

 

Regards

Edited by Saml, 16 August 2017 - 04:34 PM.