Seeking help from math wizards untangling totalizer errors

[Bit_Bucket_07]

I was called into a plant recently to diagnose errors with their natural gas totalization. Long story short: They had installed orifice plate differential pressure meters and didn't understand that they needed to extract a square root value from the analog input from these transmitters in order to infer flow.

So, I fixed the problem and their rates and totalizations are now working correctly. I received a call this morning asking if I could help them to decipher months of recorded totalization values that they had accumulated when they were treating the DP input as though it was a linear flow value. The totalizers were being trended on an HMI, so they have this data available to them, but they don't know how to correct for the errors in the data.

I am going to work on trying to figure this out for myself, but I was hoping that someone here might have a solution that they could share with me before I manage to come up with a conversion formula on my own.

Two of the DP transmitters measure 0 to 250 inches of water, with 250 inches of water equaling 3000 SCFM of gas flow. The other meter is upstream of these two meters and measures the gas flow to both of the downstream meters. It measures 0 to 500 inches of water with 500 inches of water equaling 6000 SCFM of gas flow. Although the meters measure flow in SCFM, the HMI displays that value as SCFH, and the totalizations are done as SCFH.

Any help would be greatly appreciated.

 

[rdrast]

If the values are captured in a historical trend, it should be possible to just extract the numbers, and dump them into a spreadsheet, then do your calculation there.

 

[Bit_Bucket_07]

If the values are captured in a historical trend, it should be possible to just extract the numbers, and dump them into a spreadsheet, then do your calculation there.

Sure, I can convert the trend files to csv files. It is Wonderware InTouch HMI.

I just don't know how to manipulate the data in order to correct for the totalization errors. I'm not even sure whether this is feasible.

 

[TWS]

Is this an orifice plate or flow tube? The customer should have calibration data for the primary devices.
This might help:

http://www.instrumentationtoolbox.com/2014/09/aga-3-gas-flow-equation-for-orifice.html#axzz3YuF2iFna.

Hope you get paid by the hour

 

[Bit_Bucket_07]

Is this an orifice plate or flow tube? The customer should have calibration data for the primary devices.
This might help:

http://www.instrumentationtoolbox.com/2014/09/aga-3-gas-flow-equation-for-orifice.html#axzz3YuF2iFna.

Hope you get paid by the hour

As I said, the problem has now been corrected. I don't need any help in that regard. The client had the data sheets for the DP meters.

What I am seeking is a method by which the erroneous data that was recorded prior to correcting the problem can be linearized. I could easily linearize the erroneous flow rates that existed prior to fixing the problem, but I can't seem to get a handle on whether the erroneous historical totalization data can be corrected and made useful to my client.

 

[Brian123]

If I were you, I would calculate the average flow rate that had to happen to move the totalizer from the number recorded at timestamp n to timestamp n+1. Do that for all of your timestamp intervals. Then correct the average interval flows to the actual flow rate that would have been happening. After that, totalize the new flow rate. Hopefully the interval isn't too big and the flow didn't vary to much within the interval..

 

[Bit_Bucket_07]

If I were you, I would calculate the average flow rate that had to happen to move the totalizer from the number recorded at timestamp n to timestamp n+1. Do that for all of your timestamp intervals. Then correct the average interval flows to the actual flow rate that would have been happening. After that, totalize the new flow rate. Hopefully the interval isn't too big and the flow didn't vary to much within the interval..

That's sort of what I've been thinking. I don't believe that it's possible to correct their historical data without basically re-totalizing it. That said, I'm not really a mathematical genius, and I was hoping that someone might know a trick of which I'm unaware.

 

[rdrast]

The values in the trend are the as recorded values (perhaps with a linear scaling factor applied?) of the instrument without the square root compensation, right?

So, put them in a column in Excel, in the next column, the result is the square root of the previous column's cell. B2 = SQRT(A2) Copy that formula down to all cells in column B. Now you have your normalized (flow) values.

If you need to completely re-totalize, you can use your now normalized flow values, and the time's of the samples in the trend.

 

[Bit_Bucket_07]

The values in the trend are the as recorded values (perhaps with a linear scaling factor applied?) of the instrument without the square root compensation, right?

So, put them in a column in Excel, in the next column, the result is the square root of the previous column's cell. B2 = SQRT(A2) Copy that formula down to all cells in column B. Now you have your normalized (flow) values.

If you need to completely re-totalize, you can use your now normalized flow values, and the time's of the samples in the trend.

It's not quite as simple as finding the square root of the un-linearized flow rates. Besides, I don't know for certain whether those flow rates were even trended. They probably were.

I'm about ready to tell them that correcting and re-totalizing either the historical flow rates, or the inferred rates derived from the method that Brian123 recommended is the only solution, and that their totalization trends cannot be corrected otherwise.

 

[Peter Nachtwey]

The totalizers were being trended on an HMI, so they have this data available to them, but they don't know how to correct for the errors in the data.

This is sad.

Two of the DP transmitters measure 0 to 250 inches of water, with 250 inches of water equaling 3000 SCFM of gas flow.

Flow=3000*sqrt(inches_of_water/250)

The other meter is upstream of these two meters and measures the gas flow to both of the downstream meters. It measures 0 to 500 inches of water with 500 inches of water equaling 6000 SCFM of gas flow.

flow=6000*sqrt(inches_of_water/500)
flow is in SCFM. If you want SCFH multiply by 60.

 

[Bit_Bucket_07]

This is sad.


Flow=3000*sqrt(inches_of_water/250)


flow=6000*sqrt(inches_of_water/500)
flow is in SCFM. If you want SCFH multiply by 60.

Yes. As I said, I've already corrected the problem.

I've also called and told them that the only recourse for calculating their gas usage prior to my having corrected the problem is to re-totalize based upon linearization of the old flow rates.

 

[danw]

If I understand this, the recorded data is totalizer-over-time. If you subtract a successive value from the previous value, you get a total over delta T (time), say 135 gallons over 10 seconds. Extrapolate that to a per minute value, 810 gpm (135 * 6 gpm), the 'average flow rate for that 10 second period. That can be back-calculated to correct for false reporting of percentage of DP, rather than the square root of the percentage DP.

When back calculating, don't make the mistake some do, and use the representative flow rate values, those in gpm, lpm, gph, lph, CFM or CFH values.

Differential pressure flow measurement has a square root function, but the square root function is never applied to the representative values associated with the differential pressure.

Rather, the square root function is applied to the normalized percentage of DP, not to the value 100% DP represents, say 15,000 gpm or 400 lpm.

In one application, a 100"wc DP might represent 15,000 gpm. In another application, 100"wc DP might represent 400 lpm. Both are DP flow applications but the square root of 15,000 is nowhere near the square root of 400. However, the square root of 100% or 1.00 (or any percentage of raw DP) is the same in both cases.

Column A is % raw DP, as read from transmitter, where 100% DP is the value on the sizing sheet for max flow rate. (100 inches wc in my example above)

Column B is the normalized decimal value of the raw DP percent value in column A, % DP. 64% = 0.64

Column C is the Square Root of column B. sq rt of 0.64 = 0.800

Column D is the normalized decimal value in percent of the flow rate. 0.800 = 80% of the max flow rate

The max flow rate (at 100% DP, 15,000 gpm in my example), is multipled by column C (64% DP or 0.64) to get the flow rate in its flow units 15,000 * 0.64 = 9,600 gpm.

To back calculate a falsely reported flow rate based on raw DP, not sq rt of (normalized percentage) DP, you need to know the max flow rate, 15,000 gpm.

For example, the reported false value 9600 gpm is ratioed to find its percentage of max flow: 9600/15000 = 0.64 = 64.0%
That is the percentage of raw DP, 64.0%.

Take the square root of 64% (of the raw DP) = sq rt 0.64 = 0.800 = 80.0% of max flow rate.

The true flow rate is 80% of max flow rate, or 15,000 * 0.80 = 12,000 gpm.

Dan Weise

 

[osmanmom]

Hi Danw,

Any tips how to get flow error from DP transmitter,

i have attach a file but i don't understand GCF,any tips,

thanks in advance