GPA TP-31-2014 GPA 2261 and GPA 2177 Methods Precision Statements Calculation.pdf

1、Technical PublicationTP-31GPA 2261 and GPA 2177 Methods Precision Statements CalculationRicardo AguiarMovilab, S.A. de C.V.Naucalpan, Edo. MexicoApril 2014GasProcessorsAssociation 6526 East 60th Street Tulsa, Oklahoma 74145 Phone: 918/493-3872 GPAglobal.org1 GPA Disclaimer GPA publications necessari

2、ly address problems of a general nature and may be used by anyone desiring to do so. Every effort has been made by GPA to assure accuracy and reliability of the information contained in its publications. With respect to particular circumstances, local, state, and federal laws and regulations should

3、be reviewed. It is not the intent of GPA to assume the duties of employers, manufacturers, or suppliers to warn and properly train employees, or others exposed, concerning health and safety risks or precautions. GPA makes no representation, warranty, or guarantee in connection with this publication

4、and hereby expressly disclaims any liability or responsibility for loss or damage resulting from its use or for the violation of any federal, state, or municipal regulation with which this publication may conflict, or for any infringement of letters of patent regarding apparatus, equipment, or metho

5、d so covered. Forward The “Precision Criteria” derived from the work behind RR-188 and the data regression described in TP-31 may be significantly tighter in many cases than what was listed in previous revisions of GPA 2261 and GPA 2177. Therefore, the determination of what constitutes a “Pass/Fail”

6、 may need to be based on more than a single component failing to meet the precision criteria (depending on the degree of failure). It is feasible that an instrument could fail to meet the new “Precision Criteria”, yet produce analytical results that are acceptable otherwise. Agreements between parti

7、es should consider this and determine “Pass/Fail” based on actual impact of the instrument performance. Some parameters other than concentration for consideration are Heating Value, Relative Density and Theoretical Hydrocarbon Liquid Content. “Copyright2014 by Gas Processors Association. All rights

8、reserved. No part of this Report may be reproduced without the written consent of the Gas Processors Association.“ 2 TECHNICAL PAPER TP-31 GPA 2261 AND GPA 2177 METHODS PRECISION STATEMENTS CALCULATION 1. INTRODUCTION. This technical paper summarizes calculations for GPA 2261 and GPA 2177 methods pr

9、ecision statements, repeatability and reproducibility from data obtained during Round Robin RR 188. The interlaboratory study design for GPA 2261 consisted of 10 samples with 10 components covering the application range. Such samples were analyzed in duplicate by six laboratories, resulting in a 60

10、pair data set. The GPA 2277 design consisted of six samples including 10 components analyzed in duplicate by six laboratories, resulting in a 36 pair data set. Two international standards were selected for data analysis: ISO 5725-2:94 “Accuracy (trueness and precision) of measurement methods and res

11、ults Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method”, and ASTM D6300-08 “Standard Practice for Determination of Precision and Bias Data for Use in Test Methods for Petroleum Products and Lubricants”. This Practice differs slightly fro

12、m related portions of the ISO 4259:06 standard, “Petroleum products Determination and application of precision data in relation to methods of test”. Both procedures include sample outlier identification, as well as rejection and calculation of precision estimates. The ASTM procedure requests at leas

13、t 30 degrees of freedom. That means 30 pairs of repeatability results covered in the RR 188. Some differences between selected procedures are: - In the ASTM D6300 (ASTM) procedure, identification and rejection of sample outliers come after a function transformation; therefore, in some cases, both pr

14、ocedures present different data to be rejected. - For precision estimate calculation, both methods use different weighting for regression. ASTM forces obtaining of the same exponent for repeatability and reproducibility; nevertheless, in the case of natural gas, only four out of 10 components have a

15、 dispersion profile that fits this condition. In the other cases, the repeatability function is forced to the exponent obtained in the estimated function of ISO 5725-2 (ISO). - ISO produces three estimates: a) linear zero origin, b) linear and c) exponential. ASTM is not limited to a specific functi

16、on, but produces a unique function with the same exponent and different coefficients for repeatability and reproducibility. When the repeatability function is 3 not adequate, a new regression for repeatability is performed forced with the founded ISO repeatability exponent. These differences in proc

17、edures yielded different estimates. Many of them have very tight overlapping; however, only in some cases significant differences requiring intervention were found. In those cases the rule applied was selection of the less strict function. Precision calculations were performed with a commercial spre

18、adsheet. For purposes of validation of calculation sheets, example 3 of method ISO 5725-2:1994 and data from Table A.2.1 of the ASTM D 6300-08 method were calculated with satisfactory results. For outlier identification purposes, both procedures used Grubbs, Cochrans and/or Mandels approaches. In fe

19、w cases, rejection of more than one or two values was necessary. At some level of rejection or inclusion of outliers no appreciable effect was observed because these data fell outside precision function obtained. For data rejection purposes, one of many considerations taken is the fact that when a r

20、eproducibility outlier is found in one component, such data should be rejected for all the components due to the normalization process causing a data set for a sample to be dependent upon components. This is true when concentration of that component is large enough to affect the total composition. D

21、ata were not rejected in cases where influence of a component outlier did not produce significant difference against a robust mean. This consideration was based upon the robustness of the normalization process applied to the analytical results that compensated the bias for a specific component in th

22、e whole composition. A deviation from the GPA 2261 standard was found because the sum of raw concentration (external standard calculations) reported by some laboratories did not comply with Note 14 of the method. Only 66% (2 out of 3) fell from 99% to 101% of the not normalized total, and 75% (3 out

23、 of 4) fell from 98% to 102%. This deviation may be attributed to the concentration scope of the Round Robin which exceeded the limits of concentration for both methods. For data analysis reported for participant laboratories, outliers and stragglers were identified through Mandels h and k statistic

24、s. Graphical representation thereof is shown in graphs 1 to 4. Graphs 1 and 3 show number k statistics, related to repeatability, before outlier rejection for each single component. Graphs 2 and 4 show reproducibility related Mandels h statistics, in which suspected outliers are easily identified in

25、 bars exceeding the 5% limits. 4 2. GPA 2261 PRECISION RESULTS Following previous GPA standards, repeatability and reproducibility statements are established on an individual basis. After reviewing resulting functions, only those in cxd form were considered and summarized in Table 2.1, and shown in

26、graph 2. 2.1 Component remarks. Methane reproducibility and repeatability functions displayed different behaviors because Methane has better reproducibility as concentration approaches 100%; repeatability, however, increases proportionally with concentration. For this reason the ASTM r function was

27、calculated with the r ISO exponent. Even when concentration range for methane was defined from 28.9 mole percent to almost 100%, the reproducibility function only covered 50 to 100 mole percent and cannot be applied by extrapolation to lower concentrations. Upon eliminating all Laboratory 1 i-Butane

28、 reproducibility data sets in the ASTM model, profiles of ISO and ASTM got closer, although the ASTM interval was greater. The ASTM repeatability function was fitted to the ISO exponent with an almost identical profile; in the case of ISO, however, all the repeatability data sets from laboratory 1 w

29、ere eliminated, while in ASTM only samples 5 and 6 were eliminated. As ASTM had fewer reproducibility outliers for Nitrogen, its function grew wider than the ISO function. Elimination of repeatability data for sample 7 of Lab 2 and sample 9 of Lab 2 (rejected as ISO repeatability outliers), caused b

30、oth r functions to become very similar. Carbon Dioxide ASTM reproducibility outliers increased to include samples 7, 8 and 9 of Lab 5. Removing gave rise to a behavior very similar to that of the ISO reproducibility function. 5 Table 2.1 GPA 2261 Component ISO 5725-2 ASTM D6300 Methane r 0.0106 X0.2

31、99 0.0079 X1/3 R 27236 X-2.8 91000 X-3 Ethane r 0.0102 X0.346 0.0124 X1/3 R 0.0192 X0.289 0.0315 X1/3 Propane r 0.0070 X0.148 0.0084 X1/8 R 0.0242 X0.547 0.026 X1/2 i-Butane r 0.0054 X0.190 0.0053 X1/5 R 0.0178 X0.514 0.0254 X1/2 n-Butane r 0.0118 X0.395 0.0081 X2/5 R 0.0323 X0.518 0.0327 X1/2 i-Pen

32、tane r 0.0073 X0.018 0.009 X1/4 R 0.0235 X0.248 0.0245 X1/4 n-Pentane r 0.0100 X0.201 0.0075 X1/5 R 0.0193 X0.305 0.026 X1/3 C6+ r 0.0136 X0.409 0.0135 X2/5 R 0.0514 X0.496 0.0574 X1/2 Nitrogen r 0.0403 X0.259 0.039 X1/4 R 0.174 X0.582 0.158 X1/2 Carbon Dioxide r 0.0061 X0.300 0.0042 X1/3 R 0.11 X0.

33、441 0.12 X1/3 3. GPA 2177 PRECISION RESULTS Following previous GPA 2177 repeatability and reproducibility statements, this paper presents a single reproducibility and repeatability function for all components of the analysis. After reviewing resulting functions only those in the cxd form were consid

34、ered and are summarized in tables 3.1 and 3.2. The final function for the reproducibility statement was obtained from the Propane ASTM reproducibility function, since this covers the whole range of concentrations, has a very similar GPA 2177-03 behavior and a value greater than other components, exc

35、ept Ethane and C6+ at high concentrations. Graph 3.1 shows the stair shape of GPA 2177-03 reproducibility statement and continuous functions obtained by regression. Repeatability was defined by a calculated function covering the whole range of concentrations considering calculated estimates by ISO 5

36、725-2:94 and ASTM D6300-08, as observed in graph 3.1. The goal of this function was to obtain the right value including all component repeatability functions, within the full range of concentrations, particularly ISO Nitrogen function and ASTM Propane function at concentrations below 1 %mole and C6+

37、, at high values. Ethane ISO function with coefficient of 0.08, instead of 0.03, yielded a function (0.08x2/7) covering these conditions; moreover, this function reasonably matched the GPA 2177-03 stair shape repeatability, as observed in graph 3.1. 6 3.2 Precision Results Table 3.1 GPA 2177 Reprodu

38、cibility Component ISO ASTM Methane 0.0713 x0.519 0.0726 x1/2 Ethane 0.156 x0.440 0.1427 x3/8 Propane 0.103 x0.348 0.1233 x1/3 i-Butane 0.121 x0.104 0.1261 x1/16 n-Butane 0.086 x0.329 0.0767 x1/3 i-Pentane 0.0523 x0.273 0.0630 x1/4 n-Pentane 0.0725 x0.255 0.0642 x1/4 C6+ 0.135 x0.462 0.1220 x1/2 Nit

39、rogen 0.089 x0.459 0.0962 x1/2 Carbon Dioxide 0.115 x0.413 0.1018 x2/5 Table 3.2 GPA 2177 Repeatability Component ISO 5725-2:94 ASTM D6300-08 Methane 0.0236 x0.403 0.0408 x1/2 Ethane 0.0299 x0.284 0.0300 x3/8 Propane 0.0355 x0.227 0.0573 x1/6 i-Butane 0.0302 x0.0035 0.0309 x1/16 n-Butane 0.0254 x0.2

40、75 0.0267 x1/3 i-Pentane 0.0286 x0.312 0.0259 x1/4 n-Pentane 0.0246 x0.362 0.0412 x1/4 C6+ 0.0315 x0.545 0.0356 x1/2 Nitrogen 0.0463 x0.703 0.0431 x1/2 Carbon Dioxide 0.0175 x0.381 0.0210 x3/8 4. PRECISION STATEMENT PROPOSALS 4.1 GPA 2261. The final precision proposal for GPA 2261 is shown in Table

41、4.1. Table 4.1 GPA 2261 Precision Statement Component Range Repeatability Reproducibility Nitrogen .02 15 0.039 X1/4 0.158 X1/2 Methane 50 100 0.0079 X1/3 91000 X-3 Carbon Dioxide .02 15 0.0042 X1/3 0.12 X1/3 Ethane .02 15 0.0124 X1/3 0.0315 X1/3 Propane .02 15 0.0084 X1/8 0.026 X1/2 iso-Butane .02

42、8 0.01 X1/5 0.018 X1/2 n-Butane .02 8 0.0117 X2/5 0.033 X1/2 iso-Pentane .02 4 0.009 X1/4 0.025 X1/4 n-Pentane .02 4 0.01 X1/5 0.026 X1/3 Hexanes Plus .02 2 0.0135 X2/5 0.051 X1/2 7 4.2 GPA 2177. The final precision proposal for GPA 2261 is presented in table 4.2. Table 4.2 GPA 2177 Precision Statem

43、ent Repeatability 0.08 X2/7 Reproducibility 0.1233 X1/3 5. CONCLUSIONS A new precision statement obtained from Round Robin data from the RR-188 interlaboratory program is submitted as a proposal. A selection from two procedures for calculations was useful to compare, adjust or correct both results.

44、Selection of final expressions or precision functions was conducted upon consensus with members of the working group of section B of GPA, and submitted as a proposal. This paper represents a valuable improvement for precision statements for international natural gas chromatographic methods, since cu

45、rrent laboratory proficiency schemes lack reliable reproducibility statements for laboratory performance qualification. Nitrogen-3-2-10123123456Laboratory iMandels statistic,hCO2-3-2-10123123456Laboratory iMandels statistic,hMethane-3-2-10123123456Laboratory iMandels statistic,hEthane-3-2-1012312345

46、6Laboratory iMandels statistic,hPropane-3-2-10123123456Laboratory iMandels statistic,hi-Butane-3-2-10123123456Laboratory iMandels statistic,hn-Butane-3-2-10123123456Laboratory iMandels statistic,hi-Pentane-3-2-10123123456Laboratory iMandels statistic,hn-Pentane-3-2-10123123456Laboratory iMandels sta

47、tistic,hC6+-3-2-10123123456Laboratory iMandels statistic,hGrapic 1.4 GPA 2177 Mandels between-laboratory consistency statistic, hNitrogen0123123456Laboratory iMandels statistic,kCO20123123456Laboratory iMandels statistic,kMethane0123123456Laboratory iMandels statistic,kEthane0123123456Laboratory iMa

48、ndels statistic,kPropane0123123456Laboratory iMandels statistic,ki-Butane0123123456Laboratory iMandels statistic,kn-Butane0123123456Laboratory iMandels statistic,ki-Pentane0123123456Laboratory iMandels statistic,kn-Pentane0123123456Laboratory iMandels statistic,kC6+0123123456Laboratory iMandels stat

49、istic,kGrapic 1.3 GPA 2177 Mandels within-laboratory consistency statistic, kn-Pentane-3-2-10123123456Laboratory iMandels statistic,hNitrogen-3-2-10123123456Laboratory iMandels statistic,hEthane-3-2-10123123456Laboratory iMandels statistic,hMethane-3-2-10123123456Laboratory iMandels statistic,hPropane-3-2-10123123456Laboratory iMandels statistic,hi-Pentane-3-2-10123123456Laboratory iMandels statistic,hC6+-3-2-10123123456Laboratory iMandels statistic,hn-Butane-3-2-10123123456Laboratory iMandels statistic,hCO2-3-2-101231234