You are on page 1of 13

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH 1.

INTRODUCTION: The measurement of flow through closed pipe is of most importance in our industry. Flow measurements are necessary to determine the material balance of a unit, amount of product delivered etc and for optimization of individual units or complex. Obviously, flow should be measured with as high accuracy as possible. Various types of flow meters are available, selection of which is determined based on process fluid conditions, accuracy, rangeability requirements, upstream pipe requirements and cost. Characteristics of commonly used meters are given in Table 1. This paper presents various sources of errors in measurement. The effect of each on overall measurement uncertainty is discussed so that effort can be concentrated on reducing those, which have most effect on flow metering uncertainty. In the Jamnagar complex we have following types of flow meters: 1. Orifice Meters 2. Annubars 3. Mass flow meters 4. Turbine Meters 5. Vortex Meters 6. Magnetic flow meters The Orifice Meters constitute majority followed by Annubars and Mass flow meters. Discussion in this paper is mainly on orifice meters and stress is given to application in liquid service. 2. DEFINITION OF UNCERTAINTY ISO 5167 defines uncertainty as Range of values within which the true value of measurement is estimated to lie at the 95% probability level 3. UNCERTAINITY IN ORIFICE METERS DP type meters are less expensive, simpler and more flexible devices. For these reasons, great many investigations were carried out to determine characteristics of these meters. As a result, it is now possible to predict the meter performance by calculation. Often, the meters secondary instrument gets all attention while errors caused by the primary element is ignored. The errors caused by primary elements can be classified as 1. Uncertainties linked to the quantities specified in ISO 5167 - FOSSIL ERRORS 2. Uncertainties linked to the measurements of d and D 3. Uncertainties linked to change in operating conditions 4. Uncertainties linked to Imprecision in installation- Fixed errors 5. Uncertainties linked to Imprecision in installation-Transient errors 3.1 The Basic Flow Equation The volume flow derived from Bernoullis equation is Qv = /4 * d2 * 2DP/ * 1/ 1- 4 1/13

However, there are a few factors, which influence this equation i.e.
UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

By: Pramod Kumar, CES Instrumentation

There may be significant friction in the pipe The constriction of the fluid stream is determined by the vena contracta and so is the average velocity in the pipe It may not be possible to measure the pressure exactly at the point of obstruction (orifice hole). thus the pressure at some other point in the vicinity of constriction has to be measured There are suction or impact effect at the pressure taps holes

To compensate for these factors, an empirical discharge coefficient (C) is used. C = Actual Flow/ Theoretical Flow

Thus, the actual working equation for liquids changes to Qv = /4 * d2 * C* 2DP/ * 1/ 1- 4 Qm = /4 * d2 * C* 2DP* / 1/ 1- 4

and for gases, Qv Qm = /4 * d2 * C* *2DP/ * 1/ 1- 4 = /4 * d2 * C* * 2DP* / 1/ 1- 4

3.2 Practical computation of errors: To carry out a rigorous analysis of uncertainty in any measurement, it is necessary to identify the independent sources of error, which contribute to the error in the overall measurement. When a variable y cannot be measured directly, but is a function of several other variables (x1,x2, x3xn) and its value to be computed from measurements of them, then uncertainty in the value obtained, Ey, is given by Ey2 = (dy/dx1)2 e2(x1) + (dy/dx2)2 e2(x2) + (dy/dx3)2 e2(x3) + + (dy/dxn)2 e2(xn)

Where dy/dx1 is partial derivative of y w.r.t. variable x1 In the formula for Qm also, the various quantities listed are not independent, so it is not correct to compute the uncertainty of Qm directly from the uncertainties of these quantities. However, it is sufficient for most practical purpose to assume that the uncertainties of C, , , independent of each other.

d, DP are

A practical working formula for dQm may then be derived, which takes into account the interdependence of C on d and D which E2qm = E2c + E 2 + [24/ C(1- 4)] 2 ED2 + 4 [1+ 4/ C(1- 4)] 2 Ed2 + 1/4 E2 dp + 1/4 E2

Relative Importance of Errors The relative importance of sources of error in the five variables ( C, d, D, DP and ) above becomes apparent when a particular example is considered. Eg Orifice plate with = 0.5, corner tappings, used at pipe Re = 105 The predicted value of C is 0.6053
UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

2/13

By: Pramod Kumar, CES Instrumentation

Substituting the figures in above equation, we can write E2qm = E2c + (0.22 ED )2 + (2.22 Ed )2 + (0.5 E dp)2 + (0.5 E)2

The coefficients of the various percentage uncertainties are the "sensitivity coefficients", so called because they are the measure of the extent to which the uncertainty in one of the variable influences the overall flow rate measurement. When calculation is done based on ISO 5167, Uncertainty in C is typically 0.6 % Maximum values of uncertainties in d and D are 0.07 and 0.4 % respectively. Putting typical values for remaining two, the uncertainty table will be as follows Variable C D d DP % Uncertainity* 0.6 0.4 0.07 0.6 0.6 Sensitivity Coefficient ( E)2 1 0.360 0.22 0.008 2.22 0.024 0.5 0.090 0.5 0.090

* These are typical values which are encountered in calculations. It may be seen from the above that the major contributors are C, DP and . Hence, effort should be made to reduce errors in these three variables. In the following section, attention is focussed on the three factors C, and DP and various factors contributing to each of them. It will be seen during the course of discussion that these factors are encountered in daily operation and contribution to uncertainty in measurement is considerable. 4. ANALYSIS OF VARIOUS ERRORS 4.1 Uncertainties linked to the quantities specified in ISO 5167 - FOSSIL ERRORS The uncertainties linked to ISO 5167 are on the discharge coefficient. They give minimum uncertainty with which the measurement is unavoidably tainted since the user has no control on these values. They occur because small variations in the geometry of the device are allowed and the investigations on which the values have been based could not be made under ideal conditions, nor with out some uncertainty. These errors are termed as FOSSIL ERRORS Fossil error in C For all three types of tappings, when , D, Re are known with out error, the relative uncertainty of the value of C is equal to For 0.6 For 0.6 < 0.75 0.6% same as the value of expressed in %

4.2 ERROR DUE TO CHANGE IN DISCHARGE COEFFICIENT As discussed earlier,


UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

3/13

By: Pramod Kumar, CES Instrumentation

C = ACTUAL FLOW (Qa) / CALCULATED FLOW (Qc) Qc = Qa / C Qc 1/ C When C increases, Qc decreases, calculated flow will be less if C decreases. C is assumed to be constant in calculation. A change in C will reflect as change in flow calculated. The discharge coefficient is a function of several variables viz. 1. ratio 2. Reynolds Number 3. Location of pressure tapping 4. Pipe Roughness If we assume that the orifice plate is properly made and well installed, then we may write C = (Re, , Pressure tapping, Pipe roughness) Out of the four parameters above, the last three are function of orifice installation and can have only fixed error for a given installation. We can write C = (Re) for a given installation

Discharge Coefficient as function of Reynolds Number Fig 1 shows value of C as a function of Re for different and pipe dia Inference: 1. Value of C remains practically constant above a certain Reynolds number, depending on pipe dia and ratio. 2. Lower limit of pipe Re have been established for different pipe sizes and . 3. Above these values, change in flow rate (and the Re ) will have negligible effect on value of C and makes correct measurement of flow possible. [The upper limit of Re is not probably not limited. ISO 5167 specifies Re upper value as 108 as no experimental data is available above this.] 4. As long as flow is turbulent, change in viscocity does not affect flow measurement accuracy. 5. Reynolds number effect on meter accuracy depends on ratio. In general, larger the ratio, the more critical the Re. 6. For ratio of 0.5, Re of 10000 may lead to errors in flow measurement up to 2 to 3% of range. For ratio of 0.7, the errors may be as high as 7 to 8%. 7. At low low flow rate, repeatable, controllable measurement is achievable, but accuracy at lower end may be sacrificed unless special precautions are not taken. A case study of metering uncertainty at Low Reynolds number caused by Low flow rate is presented in Annexure 1. 4.3 ERROR DUE TO DENSITY CHANGES The value of density used in calculation affect the metering uncertainties. To carry out accurate measurement, it is necessary to know the density at operating conditions. The density figures used in orifice design can vary with actual process conditions. This can be due to 1. Actual change in process fluid and its density - Online compensation is not possible UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH 4/13 By: Pramod Kumar, CES Instrumentation

2. Density changes with change in line temperature - Online compensation is possible. Actual change in process fluid and its density In case all other parameters remain constant and only density varies beyond +/- 1% of design value, an additional inaccuracy of +/- 0.5% will be encountered. +/- X% change in density introduces +/- X/2 % change in flow measured. Density Compensation for change in temperature The online measurement of these parameters and use of them in DCS calculations can ensure that the difference between the design conditions considered for the flow measuring device / instrument and the actual line conditions are monitored and flowmeter readings are accordingly corrected. As per the UOP metering philosophy / the P &IDS the Pressure & Temperature compensation for gases and Temperature compensation for liquid has been applied for the flow integrators whereas for indication / control the uncompensated flowmeter reading is utilised. The equation utilised in the DCS is given below. Liquid Flow Compensation function to compensate the flow for changes in operating temperature and specific gravity. Output = Flow x SQRT { [(d b)/(Td Tb)] x [(Tf Tb)/ Where Td

d] + [(b/d)] }

d
Tb Tf

is the design temperature in oC or oF is the design specific gravity is the base temperature (15 oC or 60 oF ) is the base specific gravity is the flowing temperature in oC or oF

4.4 Error due to DP transmitter Error The significant quantities, which affect DP transmitter, are ambient temperature effects, static pressure effects, long term drift, hysterisis, repeatability and the calibration standard uncertainty. With Smart transmitters, the error in DP measurement is much less and can be assumed to be about +/- 0.2%. This error can be kept with in limit by regular calibration. 5. Uncertainties linked to Imprecision in Installation- Fixed errors These errors can be classified as Pipe Effects and Plate Effects 5.1 Pipe Effects: The three pipe effects are Swirling flow, Circularity of Pipe, Pipe Misalignment and Pipe roughness Straight Run Requirement The orifice plate should preferably be installed at a point where there is no turbulence. In practice, bends, partly open valves etc upstream of orifice assembly tend to change velocity pattern into non-uniform one. This results in an impact effect, which causes DP to increase or decrease and hence causes meter reading to read high or low. To avoid disturbance, minimum straight run is recommended as shown in Table 2 When the upstream or down stream straight run is shorter than the zero additional uncertainty, and equal to or greater than 0.5% additional uncertainty values, an additional uncertainty of 0.5% shall be added arithmetically to the uncertainty in C. If the straight runs are shorter than the 0.5% additional uncertainty values specified, ISO 5167 specifies no information to predict the value of any additional uncertainty to be taken into account. This is also the case
UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

5/13

By: Pramod Kumar, CES Instrumentation

when the upstream and down stream straight run is shorter than the zero additional uncertainty, and equal to or greater than 0.5% additional uncertainty values. Circularity of Pipe As per ISO 5167, in the immediate vicinity of the primary device the following requirements shall apply. The length of the upstream pipe section adjacent to the primary device shall be at least 2D and cylindrical. The pipe is said to be cylindrical when no diameter in any plane differs by more than 0.3% from the mean value of D. Beyond 2D from the primary device in the upstream pipe run between the primary device and the first upstream fitting or disturbance may be made up more than one section of pipe. No additional uncertainty is involved provided that the diameter step between the two sections does not exceed 0.3% of mean value of D. An additional uncertainty of 0.2% shall be arithmetically added to uncertainty for C if diameter step D between any two sections exceeds the limit given above but complies with following relationship D/D 0.002 [(s/D + 0.4) / 0.1 + 2.34 ] and D/D 0.05 where s is the distance of the step from the upstream pressure tapping. Misalignment Out of alignment pipe can cause errors of several percent depending on which way the tube is misaligned. It is important to check gasket protrusion into the pipe interior. Pipe Roughness Even for devices, which are similar in geometry and at the same Re, the C is found to vary slightly with the device. These variations may be attributed to roughness in the upstream pipe wall relative to pipe diameter. In turbulent flow, there is a laminar boundary layer in the pipe, which depends on the Re. When Re is low, there is only a turbulent core and a rather thick boundary layer across the wall. With the increasing value of Re, the thickness of laminar layer progressively decreases. Normal pipes will behave as smooth as long as laminar boundary layer thickness is greater than the height of the irregularities on the pipe wall, but will become rough for greater values of Re when layer is thinner and irregularities project to the turbulent core. The size of the surface granules, average peak to hollow height at the surface is called roughness, k, expressed in mm. Since direct measurement of k is not possible, following assumptions will be made in calculation: Commercial steel pipes of same size are having equal roughness The absolute size of k is constant.

As per AGA, part 2 average pipe wall roughness for for ratio less than 0.6 must not exceed 300 microinch. It is further stated that the average roughness for 0.6 and greater beta ratio be no more than 250 microinch. For ratio gretaer than or equal to 0.55, and an average tube roughness of 150 micro inches or greater, flow can be under registered by as much as 1%. 5.2 Plate Effects The major plate effects are Orifice Plate Eccentricity, Nicked Orifice Plate Edge, Metering tap Obstruction, Bent Plate, Reversed Beveled Plate
UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

6/13

By: Pramod Kumar, CES Instrumentation

Orifice Plate Eccentricity As per ISO 5167, the primary device shall be perpendicular to the Centre line of the pipe to within 10 The distance ex between the center line of the orifice and center lines of the pipe on upstream and down stream sides shall be less than or equal to 0.0025D / 0.1+2.34 If 0.0025D / 0.1+2.34 < ex < 0.005D / 0.1+2.34 , an additional uncertainty of 0.3% shall be added arithmetically to discharge coefficient C. In case where ex > 0.005 D/ 0.1+2.34 ISO 5167 does not provide any information by which to predict additional uncertainty to be taken into account. Nicked Orifice Plate Edge Tests have shown that plate that does not have a sharp upstream edge can produce between a -1% and -13% metering error. Metering tap Obstruction Debris collected on tap can cause errors on higher or lower side depending on the tapping. Bent plate Orifice plate usually happens when DP across plate exceeds. Plates deformed in flow direction can cause negative measurement errors. Reverse Beveled Plate Beveled orifice plates incorrectly installed with bevel upstream will produce significant under- registration of flow. Errors of 9 % 20% are expected. Pipe effects and plate effects can be overcome by the use of calibrated meter run. It has Machined internals Orifice flanges welded to the pipe in a workshop Upstream and downstream length are exactly as per standard Pressure tapings are smooth and at designated location as per standard Advised where greater accuracy/ custody transfer is required. 6 Uncertainties linked to Imprecision in installation-Transient errors Transient errors are caused because of (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix) Improper laying of the impulse tubing avoid improper slope, sharp bends. Leakage in the impulse line / root valves. Congealing of liquid in impulse lines Improper steam tracing Water condensate in LP/ HP tapping Condensate pots for LP & HP tappings are not installed at the same elevation. Improper installation of flow transmitter/ impulse tubing transmitters for liquid and steam service should be mounted below the pipeline and in case of gas service , they should be mounted above the pipeline. Equalizing valve passing. Transmitter drain valve leakage

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

7/13

By: Pramod Kumar, CES Instrumentation

7. Mismatch in Readings of Two Meters installed on two sections of the line. Very often it is observed that two meters installed in two plants on the same product line are not matching due to reasons listed above i.e. 1. 2. 3. 4. Pipe effects Plate effects Transient errors Mismatch in design data.

8. UNCERTAINTY ANALYSIS WHEN MORE THAN ONE METER IS USED The uncertainty analysis when a single meter is used was discussed in above section. This section discusses case when flow rate is obtained by summing up flow through more than one meter. In this case, it is not correct to arrive at the uncertainty in its measurement by simply averaging the uncertainties in individual meter. Instead, it is necessary to distinguish between the sources of random and systematic errors so that the associated uncertainties can be treated separately. Or in other words, the uncertainties are to be divided into two classes, those, which affect each meter in random manner and those, which have same effect on each meter. Uncertainty in d and D Only systematic error Uncertainty in DP Systematic component - Transmitter calibration Random component - Effect of pressure tappings etc Uncertainty in Only Systematic error Uncertainty in C Systematic error - Fossil Error Random error - Random variation in DP, Temperature, Static Pressure etc Approximation- Random and systematic components equal. Table below illustrates calculation ( the figures are indicative) Variable Effect on orifice % Uncertainty E Sensitivity Coefficient Random Systematic C R&S 0.6 0.6 1 D S 0.4 0.22 d S 0.07 2.22 DP R&S 0.5 0.5 0.5 S 1 0.5 Sum Square root

( E)2 Random 0.36 0.0625 Systematic 0.36 0.008 0.024 0.0625 0.25

0.4225 0.65

0.7045 0.839345

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

8/13

By: Pramod Kumar, CES Instrumentation

Random error component of the uncertainty in the total flow rate is 1/ 2 times random component of uncertainty in the measurement through one of the orifices, where as systematic component is equal to systematic component of individual meters. Hence uncertainty in total flow = (1/2* 0.652 + 0.70452 = 0.956 %

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

9/13

By: Pramod Kumar, CES Instrumentation

9. REFERECNCES Shell Documents: 1. Shell Flow Meter Engineering Handbook 2. Flow Measurement 16 EG- 140 3. Design Requirements for Flow Systems 16 GS- 140 Industry Standards: 1. BS 1042 / ISO 5167 - Measurement of Fluid Flow in Closed Conduits 2. ISO 5168 - Measurement of Fluid Flow- Evaluation of Uncertainties 3. API Chapter 14.3 - Concentric Square edge Orifice Meters 4. API chapter 5.1 - General Considerations for Measurement by Meters Publications 1. The Estimation of Uncertainty in Practical Situation with Orifice Meters - Kinghorn F- National Engineering Laboratory 2. Measurement Report - Hydrocrabon Processing- Witte James N- El Passo energy Company 3. Accuracy counts - How abnormal conditions affect orifice meter measurement - Thomas Morrow B Websites 1. www.spe.org 2. www.flowcontrolnetwork.com 3. www.hydrocarbonprocessing.com

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

10/13

By: Pramod Kumar, CES Instrumentation

TABLE 1- CHARATERISTICS OF COMMONLY USED METERS Type of Flowmeter Orifice Annubar Venturi Tubes Magnetic Coriolis meters Turbine Flowmeters Ultrasonic Flowmeters Vortex Variable area Meters Rangeability 3:1 3:1 3:1 10:1 20:1 10:1 Accuracy +/-1 5%URV +/-5%URV +/- 2% +/-0.5% of rate to +/-1% of URV +/-0.15 to +/0.5% of rate Gases :+/- 0.5 %of rate Liquids :+/- 1% of rate +/-1% of rate to +/-5%URV +/- 0.75 to 1.5%of rate +/- 0.5% of rate to +/-1 % of URV Permanent Pressure loss (wrt p) 40 to 80 % 5 to 10 % 5 to 20% None 5 to 20% 20 to 40 %

None 20 to 40% 5 to 20%

20:1 10:1 5:1

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

11/13

By: Pramod Kumar, CES Instrumentation

FIG 1- VARIATION OF DISCHARGE COEFFICIENT W.R.T REYNOLDS NUMBER

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

12/13

By: Pramod Kumar, CES Instrumentation

TABLE 2- STRAIGHT RUN REQUIREMENTS AS PER ISO 5167

UNCERATINITY IN FLOW METERING - A PRACTICAL APPROACH

13/13

By: Pramod Kumar, CES Instrumentation

You might also like