Flowmeter 1

[Vic123]

When sizing a turbine flowmeter for an application, what will be the accuracy of a flowmeter when say for example the vendor gives the following info:

Repeatability: ± 0.02% typical
Turndown Ratio:
(Models max. rated flow ÷ min. flow rate)
Up to 1,000:1, reduced accuracy
Up to 700:1, single fluid, constant temp.
Up to 300:1, single fluid, variable temp. and visc.
Calibrated Accuracy: ± 0.1% of rate
Linearity: ±0.15 to 0.20% of rate or better

Do we sum all the above or use any one?

 

[thermcool]

It seems the accuracy is ± 0.1% of flow rate

 

[quark]

Your question is not clear. Turndown is the ability of a meter to measure various flowrates within a given accuracy. First of all, you check your minimum and maximum flowrates and then get the accuracy of the meter within that turndown ratio.

I am new to a turbine meter that would give you an accuracy of 0.1% of measured value. Ask the manufacturer for the accuracies at 700:1 and 1000:1 also. Many times, I have seen these accuracy curves being extrapolated asymptotic to the ordinate (axis of accuracy). This is a useless case and it is better to choose a meter with comparitively lower turndowns but higher accuracies.

 

[JJPellin]

I also believe that this accuracy is going to be affected by any variance in the viscosity or specific gravity. If the numbers that you provided the meter manufacturer are not accurate, then the results from the meter will be inaccurate also. Anyone with more experience with these meters, please correct me if I am wrong.

 

[jmw]

The key to good accuracy is repeatability.
This is the ability, at any set of conditions, to obtain the same response every time you repeat the conditions.
Accuracy is then how well this value is used to interpret the actual flow rate.

Historically all meters had to be treated as “linear” because there were not the electronics to do other wise. The meter would be given a 3 point calibration to solve the equation Y=mX + c and overcheck the result. (in the case of mechanical meters “c” = O)

Turbines are generally considered to be “linear” in the upper flow ranges, that is from 20-100% of flow (turndown ratio 5:1) solving that simple equation would mean that at any flow rate within that range the accuracy would be (+/-)0.25% of reading or rate.
By extending the range to include lower flow rates and retaining the same “k”factor (pulses per unit volume) because the response is less linear, the accuracy would reduce to say (+/-)0.5% for a 10:1 turndown ratio.

For applications where the flow rate varies, the useable flow range would not be much better. In some applications it is extended to the 2.0% linearity limit, the flow rate at which the non linearity results in a lesser accuracy.

At any given flow rate while the accuracy may lie in the range (+/-)0.5%, the result will be repeatable to 0.02%. That means that at that flowrate, the reading will always have the same offset from the true accuracy.
For example, at a calibration point the accuracy would actually be the repeatability.
At another flow rate the result would always be (+) 0.4% (+/-)0.02%.

Modern electronics and automated calibration (to reduce the cost of calibration) means that “linearization” is possible. Instead a simple “Y=mX+c” solution for the whole flow range, curve fitting techniques could be employed.
Early techniques simply used the linear fit between calibration points. Thus between two calibration points the k factor will be one value and between the next two points new values for “m” and “c” are found and a new k factor value is used. The accuracy is best at the calibration points and least between them.
Curve modelling can be used to provide a good fit with fewer calibration points simply because the linear model is not necessarily the most accurate.
Because we are no longer limited to the “linear” range of the meter, we can use the entire repeatable range of the meter.

We can even define the repeatable range according to the repeatability value. If we accept a repeatability of (+/-)0.05% instead of (+/-)0.02% we can go even further down range.

The performance will now depend on the repeatability, the type of curve fit and the number and distribution of the calibration points.

Now we have to consider the other factors affecting calibration.
A turbine meter is a velocity meter.
The changing velocity of the fluid will cause the rotor to change its speed. We derive the volume from the geometry of the meter and the geometry will change with temperature and pressure. These effects are readily compensated for.

We also have to account for the effects of temperature and pressure on the fluid. These effects are well documented for hydrocarbons, for example, but less so for other fluids and if we want to get to mass flow then we also include density corrections.

One other consideration is viscosity.
Turbine meters are more viscosity sensitive than some other types of meter, straight bladed rotors more so than helical bladed.

Viscosity will change with temperature, composition, with pressure (admittedly, with hydrocarbons only significantly at high pressures and with hydrocarbons the pressure is also important to correct the density).

How viscosity correction is managed depends on a variety of factors.
Some suppliers linearise for different viscosities. The meter then uses the calibration curve for the viscosity nearest to the actual viscosity. This requires external input to “choose” the curve.
Some use the “Universal” correction factor and use a live viscosity input.
Others may plot the different viscosity curves and then use the measured viscosity to interpolate the k factor.

If viscosity varies due to temperature the viscosity may be inferred from the temperature. FTI use this approach for diesel fuel metering.

If the viscosity varies due to fluid changes or variability then a live viscosity measurement is used. Smith Meters use this approach.

Collecting all these factors together and the various corrections, it is necessary to analyse the cumulative effect of all the measurement errors and calculation errors.

A good flow computer will have multiple linearization curves, it will have temperature and pressure inputs, it will have viscosity inputs. (Don't neglect signal transmission errora: for example, a 4-20mA temperature signal has errors associated with the 4-20mA transmission and the reception.

I hope this hasn’t confused the issue still further, but may be able to understand what the supplier is saying a little better. You can now ask your supplier more specific questions.

You need to specify what you fluid is, what the process conditions are and how they change. Avoid too much “margin for error”, all too often being a bit safe/greedy can have a serious affect on meter sizing, costs and performance.
Note that temperature and pressure can have an effect both on the meter calibration and on the fluid. Density is not usually a factor in meter but it is on the fluid if you want mass flow.

JMW

http://www.ViscoAnalyser.com

 

[Vic123]

Thank you all you your help.

I guess for the info I mentioned above of the vendor, the accuracy will be 0.1 % of the rate, for a turndown ratio of 300:1 with a repeatibility 0.02%.

 

[JLSeagull]

The mention of a helical turbine raises the benefits of the Faure Herman style meters over traditional turbines.

http://www.omniflow.com/newsletters/volume1/number2.html