The MMs are 4 KRAL screw meters, normally permanently installed in the BEV test rig as secondary working standards.

Report: Intercomparison of calibration of master meters (screw meters) installed in a test rig, by two independent test methods (by BEV-standard capacity tanks and by CMI- Brooks piston prover) Euromet number 831 Authors: Libor Lojek, Gerhard Baubinder Period, in which the intercomparison took place: 07-003 until 07-004 Persons in charge: Libor Lojek c/o CMI, CZ-63800 Brno/Czech Republic, Okruzni 31 T +0040 5 45 77, E llojek@cmi.cz Gerhard Baubinder c/o BEV, A-1160 Vienna/Austria, Arltgasse 35 T +0043 1 49 110 518, E g.baubinder@metrologie.at Release date of the report: 19-09-005

1 Purpose of the project The error curves of four master meters (subsequently short MMs) (two for gasoline covering a flow rate range of 30-1400 l/min and two for diesel oil for the same flow rate range) are measured by using two different test methods: 1. use of standard capacity tanks (of the BEV test rig). use of a transportable Brooks compact piston prover "BCP" (owner CMI) connected to the master meters for the purpose of this intercomparison. By this, the reliability into the hydrocarbon CMC entries of both countries shall be reinforced. The project is comparable to BCR-project 3476/1/0/03/9/9-BCR-S(30) and anticipating on the agreed CCM.FF-K project on hydrocarbons. The MMs are 4 KRAL screw meters, normally permanently installed in the BEV test rig as secondary working standards. The deviation of the resulting CMI-error curves from the resulting BEV-error curves is a measure of the reproducibility of both test rigs. Brief history of the measurements At the first stage of the intercomparison the MMs were taken from the BEV test rig and installed in line with the BCP at CEPRO cooperation Strelice/Czech Republic. But measurements on the BEV-test rig directly following the tests at CEPRO had shown, that the vertical position of the meters in the BEV test rig with the inlet from below (see table 1 - Test specifications) caused - at certain (low) flow rates -, a pressure of the screw wheels upon the meter case, which led to a singular peak of the error curve. Subsequently, the corresponding pipe of the BEV test rig was changed so that the meters now have their inlet from above. So, a second stage of intercomparisons became necessary: contrary to the first stage the BCP came to Vienna and by this, the test liquid was identical for both the BCP and the BEV-test rig. The test results only represent the measurements of the second stage. Test rigs The CMI-test rig is a mobile Brooks Compact Prover (BCP), mounted on a trailer and used for on-site calibrations of flowmeters e.g. on product pipelines, in refineries, on the crude oil pipeline and in transmission terminals. Serial. Nr. 9609-09091-1-1, Q = 6,6 l/min 6600 l/min, tube diameter 1 inch, prover volume 60 l. It enables a dynamic measurement under operational conditions without any interruption of the measurement. Connections to the pipework are made by rubber hoses (4 m long, flanges DN100/PN16 or DN150/PN16). The liquid and the hydraulic energy is provided by external sources. The CMI-BCP is calibrated gravimetrically by the water draw method. The BEV has two stationary, equally designed test rigs (one for gasoline and one for diesel oil), each of them being equipped with two of the above mentioned KRAL screw meters as secondary working standards. The primary working standards of the BEV-test rig are static 1000 litres-standard capacity tanks for the flying and start/stop-operation (flying operation by means of a diverter), Q (test rig) = 30 l/min 1500 l/min.

The liquid and the hydraulic energy is provided by underground tanks/pumps. The standard capacity tanks are calibrated with water by delivery from a gravimetrically calibrated standard capacity tank. The MMs are calibrated by the standard capacity tanks of the test rig. 3 Meters under test The KRAL screw meters are especially suitable as master meters as well as transfer standards because they are independent of the flow profile at the meter inlet and have a good repeatability and a good long term stability (see also BCR- Kerosene - Report "Calibration Intercomparison on Flowmeters for Kerosene - Synthesis Report Part I & Part II - 3476/1/0/03/9/9-BCR-S(30) - SP, Weights and Measures Boras, 1995). Meter output by pulses (K m = 71 Imp/l for the DN40 meters and K m = 17 Imp/l for the DN100 meters); pulses fed into the electronics of the corresponding test rigs. The pulses were transmitted by one channel only, but it was assumed that no disturbances on the lines would occur, because at the comparison no source of interference was present in the vicinity of the installation (no pumps, no controllers); the good repeatability is an evidence that no spurious pulses had occurred. A long term drift is taken into consideration because of the different dates of the calibration of the MMs by the BEV-standard capacity tanks and by the CMI-BCP. As the long term drift of the error curves of the screw meters has not been determined by the BEV yet, the data for the estimation of the drift are taken from the experimental data of Kräutler, the manufacturer of the KRAL-meters (drift KRAL, see table below). screw meter drift KRAL = K/a = u long term u long terrm (% ) DN40 meters: K/a = 0,005 0,000 05 type OM5, serial Nr. B1065V (gasoline) und B1064V (diesel), K m = 71 DN100 meters: K/a = 0,00 0,000 14 type OM100, serial Nr. B681V (diesel) und B1060V (gasoline), K m = 17 Remark: The short term stability is covered by the uncertainty of repeatability.

4 Test specifications Place and date of test, used standards (first stage) first stage: CEPRO/Strelice, June 00 CMI-BCP (first and second stage) first stage: BEV/Vienna, July 00 standard capacity tanks transfer standard position of installation of meters Test liquids second stage: BEV/Vienna, July 003 standard capacity tanks; BEV/Vienna, July 004 CMI-BCP KRAL-screw meter type OM100, serial Nr. B681V, diesel oil, K = 17 Imp/l, Q = 00 000 l/min, DN100 (pipe thread connection) KRAL-screw meter type OM100, serial Nr. B1060V, gasoline, K = 17 Imp/l, Q = 00 000 l/min, DN100 KRAL-screw meter type OM5, serial Nr. B1065V, gasoline, K = 71 Imp/l, Q = 5 350 l/min, DN40 KRAL-screw meter type OM5, serial Nr. B1064V, diesel oil, K = 71 Imp/l, Q = 5 350 l/min, DN40 horizontal, coupled with outlet of BCP diesel oil ν = 4, mm /s (15 C) (first stage) gasoline ν = 0,57 mm /s (15 C) (first stage) first stage: vertical, meter inlet from below second stage: meter inlet from above diesel oil ν = 5,58 mm /s (15 C) (first stage) ν = 5,6 mm /s (15 C) (second stage) gasoline ν = 0,71 mm /s (15 C) (first stage) ν = 0,71 mm /s (15 C) (second stage) Average liquid 19 C (first stage) C (first stage) temperature * 19 C (second stage) *actual values see protocol Table 1 - Test specifications (one meter pipe thread connection DN40 and one meter pipe thread connection DN100 for gasoline, one meter pipe thread connection DN40 and one meter pipe thread connection DN100 for diesel oil). 5 Test set-up and test procedure 5.1 CEPRO/Strelice/first stage: With the BCP-tests each MM was mounted at the outlet of the BCP. The BCP was installed in a bypass of a pipeline. Prior to all test series each MM was rinsed in the installation for 5 min in order to adapt it to the liquid temperature and to separate eventually entrapped air.

The nominal flow rate was set manually (by a valve at the outlet side of the bypass and partially by a valve in the pipeline between inlet and outlet of the bypass) prior to the start of the test series for that definite flow rate. Temperature and pressure were taken from the Pt100 and pressure sensor of the BCP. The viscosity was determined by taking a sample and determining it by the BEV-lab. 5. BEV/Vienna /first stage: The MMs were then installed in the BEV-test rig in their original position. Prior to all test series each MM was rinsed in the installation for 5 min in order to adapt it to the liquid temperature and to separate eventually entrapped air. The nominal flow rate was set in a preliminary run (by a control valve at the outlet of the meters) and the adjustment of that control valve was frozen during the test series for that definite flow rate. Temperature and pressure were taken from the Pt100 and pressure sensor of the test rig. The viscosity was determined by the BEV-lab. 5.3 BEV/Vienna /second stage: The MMs were installed in the BEV-test rig in their original position. Prior to all test series each MM was rinsed in the installation for 5 min in order to adapt it to the liquid temperature and to separate eventually entrapped air. The nominal flow rate was set (by a control valve at the outlet of the meters) and the adjustment of that control valve was frozen during the test series for that definite flow rate. The pressure for the correction of the reference volume of the BCP was taken from the sensor of the BCP; on the contrary the temperature for the correction of the reference volume of the BCP was taken from the test-rig of the BEV because these temperature values seemed to be more plausible: during the tests the Pt100 of the BCP was strongly exposed to sun radiation and the Pt100 is not in direct contact with the liquid but is installed in a well so that heat transmission from outer parts might have corrupted the temperature measurements. The viscosity was determined by the BEV-lab. 6 Assessment of measuring results The test results are the error curves (K-factors at different flow rates Q j ) at a common reference temperature of 15 C. After correction of the K-factor under actual conditions to the K-factor at 15 C (see point 7 - Calculation of correction of viscosity) a least-square-fitting was performed (see point 8 - Fitting of error curve). 7 Calculation of correction of viscosity Method from Calibration Intercomparison on Flowmeters for Kerosene Synthesis Report Part I & Part II 3476/1/0/03/9/9-BCR-S(30), chapter 14 For reasons of better comparability the measurement values are given at reference conditions. These reference conditions might be (e.g. for temperature) 15 C or 0 C, but the reference temperature can also be a quite unusual value (e.g. 17,3 C). In order to keep the measurement uncertainty small by referencing the actual measurement values (by interpolation or by extrapolation) to reference conditions, that value was taken as the reference value which was approximately a mean value or which was lying in the surrounding of most of the measured conditions.

So, in the case that the test runs are performed with liquids of different temperatures/viscosities, a K-factor at reference conditions (reference temperature, reference viscosity) can be found by the following algorithm: Step 1: Temperature correction to a reference temperature T 0 (T 0 = 15 C in the present case), where only the volume expansion of the meter shell is applied to the K-factor k(t) (α being the linear expansion coefficient of the meter material). So, for a cubic expansion of the meter shell of screw meters follows: k(t 0 ) = k(t). (1 + 3α. (T T 0 ))= = k(t). (1 + 0,000035. (T T 0 )) Step : Viscosity correction for the error curves which have been adjusted acc. to step 1. When establishing the error curves for different liquids (by taking the temperaturecorrected error curves of step 1), there will be a mutual shift for reasons of the different viscosities. In order to be able to examine the dependency of the K-factors on the viscosity ν, and as a consequence to transfer the actual K-factors into K-factors at a reference viscosity ν r, the K-factors k are plotted separately for each flow rate Q i against viscosity ν. So for every Q j one yields a curve as a relation between k and ν. Based on former measurements, the following simple relations between the K-factor and the viscosity (Tab. ) have been achieved for the screw meters Viskosity ν (mm /s) 0,5 1,5 1,5,5,5 6 Relative slope ( K/1 mm /s) / K m.100 (%) + 0,19 + 0,05-0,01 Tab. Dependence of K-factor on viscosity These relative slopes K/ ν represent a mean value of all four meters. Deviations from the mean are allowed for in the calculation of the measurement uncertainty. In this step the K-factors are referenced to that viscosity which would appear at the reference temperature (in the present case of the second stage measurements 15 C). So the K-factors are valid for that liquid which would have the indicated viscosity at 15 C. 8 Fitting of error curve Method of least square fitting. Prior to the fitting, the error curve was referenced to an average viscosity (see 7 - Calculation of correction of viscosity). Errors caused by the fitting are allowed for in the calculation of uncertainties. As an assumption for the error curve the following polynomial was used: K(Q) = a 0 + a 1. Q + a.q + a 3.Q -1 + a 4.Q -

9 Measurement uncertainties MMs calibrated by standard capacity tanks of BEV-test rig (see also CMC-tables of BEV): ( 0,038 + 0,06 ) 0,1% U ( k = ) = u + er = ctestrig u cmastermet Because the measurements of the comparison had been performed in the same way as the measurements being the basis of the CMC-tables, the value U < 0,1% is still valid. Nevertheless, only a smaller number of repeated measurements could be performed during the comparison. But it can be shown by former measurements that the repeatability of the meters is extremely good and has little influence on the magnitude of the overall uncertainty. MMs calibrated by CMI-BCP The overall uncertainty U (k=) of the BCP is 0,05 % (see calibration certificate based upon CMC-database CZ8). This value takes into consideration the different ambient conditions under which a mobile prover is used. Under the assumption that the uncertainty of the screw meter being tested by the BCP is approximately the same as by the standard capacity tanks (similar test conditions like pressure, temperature, viscosity, no pressure variations, flying test method, no gaseous formations) the uncertainty of the meter can still be assessed by the same value as above (u cmastermeter = 0,06 %). So, the overall uncertainty U (k=) reads: U ( k ( u cbcp + u ) = ( 0,05 + 0,06 ) 0,08% = ) = cmaster The long term stability (expressed as the change of the K-factor per year K/a) is allowed for in the measurement uncertainty of the meter ( K/a < 0,01 for DN40 and < 0,005 for DN100). 10 Summary of measurement results Remark: The presented measurement results represent only the second stage. The meters are stable enough so that long term variations do not influence the results within the range of the given measurement uncertainty. Installation effects caused by swirls and irregular flow profiles at the meter inlet do not influence the results, because the meters are of positive displacement type, for which the flow profiles at the inlet and outlet side of the meter are irrelevant. Presentation of results 10.1 Presentation by graphical format: The measurement results are given as a fitted curve for each meter together with their uncertainty (see 4 figures below).

B1060V/DN100 =>fliegend=>bev.cmi, Juli 004, Benzin (gasoline) K-Faktor 16,840 16,835 16,830 16,85 16,80 16,815 16,810 16,805 16,800 16,795 16,790 16,785 16,780 16,775 16,770 16,765 16,760 16,755 0,1% En= 0,3 0 100 00 300 400 500 600 700 800 900 1000 1100 100 1300 1400 1500 1600 1700 Durchflussstärke BEV CMI B1065V/DN40 =>fliegend=>bev-cmi, Juli 004, Benzin (gasoline) K-Faktor 71,00 71,180 71,160 71,140 71,10 71,100 71,080 71,060 71,040 71,00 71,000 70,980 70,960 70,940 70,90 70,900 70,880 70,860 70,840 0,1% En= 0,4 0 0 40 60 80 100 10 140 160 180 00 0 40 60 80 300 30 340 360 380 400 Durchflussstärke BEV CMI

B681V/DN100 =>fliegend => BEV-CMI, Juli 004, Diesel K-Faktor 16,840 16,835 16,830 16,85 16,80 16,815 16,810 16,805 16,800 16,795 16,790 16,785 16,780 16,775 16,770 16,765 16,760 16,755 0,1% En= 0,08 0 100 00 300 400 500 600 700 800 900 1000 1100 100 1300 1400 1500 1600 1700 Durchflussstärke BEV CMI B1064V/DN40 =>fliegend=>bev-cmi, Juli 004, Diesel K-Faktor 71,00 71,180 71,160 71,140 71,10 71,100 71,080 71,060 71,040 71,00 71,000 70,980 70,960 70,940 70,90 70,900 70,880 70,860 70,840 0,1% En= 0,09 0 0 40 60 80 100 10 140 160 180 00 0 40 60 80 300 30 340 360 380 400 Durchflussstärke BEV CMI

10. Presentation according to ISO 575 (repeatability under alternated measurement conditions) In order to make the measurement results more comparable each curve is expressed by a representative K-factor, together with the repeatability and reproducibility. The differences between the error curves and between the K-factors represent the differences between the measuring results, which would be provided to a customer by using the one or the other test rig. The reproducibility of the representative K-factors is defined in accordance with ISO 575 and gives the repeatability under alternated measurement conditions. Therefore the reproducibility is the measure to assess the measurements of this comparison. The calculation of the reproducibility is based upon the following definitions:

Indices used by the definitions: I J XXX k i (Q j ) k s j j To indicate a single measurement within a series at one flowrate To distinguish a selected flowrate To identify the lab (in the present case CMI, BEV) K-factor for a single measurement i at the flowrate Q j Mean value for a series at flowrate j ki ( Q j ) k j =, 1 < i < n n n number of repeated measurements at Q j remark: n 10 (BEV), n 5 (CMI) Variance of k j from repeated measurements at flowrate Q j s j = 1/(n-1). Σ (k i - k j ) Units: (Imp/l) k(r) Representative K-factor for the present error curve 1 Q j. k j k( r) = m, 1 Q j m 1 < j < m Units: (Imp/l) m...number of selected flowrates Q j over the given flow range k(r) XXX Representative K-factor for lab XXX < k(r)> (Global) mean value of the representative K-factors of all p labs (p = in our case) < k(r)> = 1/p. Σ k(r) XXX = ½.(k(r) CMI + k(r) BEV ) Units: (Imp/l) remark: (k(r) CMI mean value from CMI Protocol, reduced to 15 C s Average calibration variance of participant (CMI / BEV), i.e. mean value for all m CMI flowrates Q s j BEV s XXXj s XXX =, 1 < j < m Units: (Imp/l) m Global mean calibration variance, i.e. average over the variances from CMI and s BEV r ec s C R ec s = 1/.( s CMI + s BEV ) Units: (Imp/l) remark : only s BEV was calculated and then set equal to CMI because the repeatability of the meters is not changed by the different standards. (inter)comparison repeatability (stating the statistical range of variances at different labs (the possible variance of any test series will lie within r ec with a 95 %-probability) 95 n rec = t 1 s Units: (Imp/l) Inter-laboratory variance, i.e. from the inter-comparison of representative K- factors between labs 1 sc = ( k( r) XXX < k( r) > ) Units: (Imp/l) 1 Reproducibility for the total (inter)comparison defining the probable maximum difference between any two laboratory concerning the representative K-factor (states the probable maximum difference of the representative K-factor between any two labs) R 95 ec tn sc + = 1 s Units: (Imp/l) s

With the above definitions the reproducibility R ec for the comparison (states the probable max. difference of the representative K-factors between CMI and BEV) is: Master Meter s B1060V k=16,8 DN100 gasoline B1065V k=71,0 DN40 gasoline B681V k=16,8 DN100 diesel oil B1064V k=71,0 DN40 diesel oil s C 95 = tn rec 1 s ReC = tn 1 sc + s (Imp/l) (Imp/l) t=3,18 for (n-1)=4 t=3,18 for (n-1)=4 Imp/l Imp/l 0,000014 0,00001 0,01 0,017 0,000081 0,00038 0,09 0,057 0,000001 0,00000 0,003 0,005 0,000337 0,00003 0,059 0,06 95

10.3 Presentation by the E n -criterion (according to EAL-P7 "EAL Interlaboratory Comparisons") As a criterion on the compatibility of two different measurements (two test rigs) of the same measurand (KRAL master meters) not only the results of measurements but also the associated uncertainties must be taken into account. The "normalized error" E n serves as a compatibility criterion with (for both labs A and B) the same fixed expansion factor k (valid for not correlated measurements as it is in the present case: two independent test rigs): x A xb 1 x A xb 1 d En = = = k U + U u + u k u( d) A B A B x A, x B results of measurement (in the present case representative K-factor k(r), see 10.) U A, U B associated expanded uncertainty, no correlation between x A and x B, x A and x B normally distributed The smaller u(d) the more reliable is the estimate d (of the true difference D = X A - X B ); but for compatibility reasons, d must also be close to zero. For k = the criterion: E 1 n is often proposed. Master Meter k(r) BEV (Imp/l) k(r) CMI (Imp/l) U BEV U CMI E n B1060V k=16,8 DN100 gasoline 16,799 16,804 + 0,10 % + 0,0168 Imp/l + 0,08 % + 0,0134 Imp/l 0,3 B1065V k=71,0 DN40 Gasolina B681V k=16,8 DN100 diesel oil B1064V k=71,0 DN40 diesel oil 71,08 71,006 + 0,10 % + 0,071 Imp/l 16,87 16,85 + 0,10 % + 0,0168 Imp/l 71,15 71,144 + 0,10 % + 0,071 Imp/l + 0,08 % + 0,057 Imp/l + 0,08 % + 0,0134 Imp/l + 0,08 % + 0,057 Imp/l 0,4 0,08 0,09