You are on page 1of 16

Soo&ty of Petroleum Engineers

SPE 30316 Pressure-Volume-Temperature Correlations for Heavy and Extra Heavy Oils

Giambattista De Ghetto*, Francesco Paone, and Marco Villa*, AGIP S.p.A. * SPE Member

fXWght

1Ss5, Sdety

Of Petmfeum Engineers, Inc. tiaavy Oil Sympmium held in Calgsry, Albarta, CenaOs, 1S-21 JunelSS5

This papar was prepared for praaametion et the Infematiwml

This peper was aektad for pmsantarion by an SPE Progrsm CommMae fallowing reviaw of inforrnatii mnteinad in an atstrsct submiied by the author(s). contents of the paper, es praaentsd by ttw sufhor(s), Tha meteriel, aa prasentad, does not neceeserily reflecf any position of the Z~*~~tk M~of Pet*~E_maMam_tim~ 4 . . . . . ...... *-. ., h e,,,., 1 ., D*,* GMi-m iti* bjj E.WI.I ..1,,,0001-..,... .-*,,... .- .. .. . . . ..a.-.. SOdOty of Petrobum Engineers, its ofkers, or members. Papars presntad ai SPiE tirrgs are aubjacf io piibbtifii Pwmiaeh to oopy is reetrktad to an abstrect of not rmra than S00 words. Ifluatrafions mey not be mpiad. The ebatrmt should mntain ooneR@uws SdmowMgmenf orwhereandtywhom Uta paper is pmaanted. Write Libmrfsn, SPE, PO. Sox SSSSSS, Riirdaon, TX 7S0SS-SS2S, U.S.A. (Facsimile 214-S52-94S5).

ABSTRACT: Thepaper

evaluates the reliability

of the most common empirical correlations used for determining

reservoir fluid properties dead-oil

whenever laboratory PVT data are not available: bubblepoint pressure, solution GOR, bubblepoint OFVF. viscosity, gas-saturated oil viscosity and undersaturated oil viscosity.

isothermal compressibility,

The reliability has been evaluated against a set of shout 65 heavy and extra-heavy oil samples. About 1200 measured data points have been collected and investigated. All measured data points are reported in the paper. For all the correlations, the following statistical parameters have been calculated: a) relative deviation between estimated and experimental values, b) average absolute percent error, c) standard deviation;
011 samples have been divided in two different API gravity ciasses: extra-heavy oils for API< The kst correlations for each class of API gravity have been evaluated for each oil-property.

10, heavy oils for 10< API<

22.3.

The functional forms of the comelations that gave the best results for each oil property have been used for finding a better comelation with errors reduced, on average. by 10%. In particular, for extra-heavy oils, since no correlations are available in literature (except for viscosity), a special investigation has been performed and new equations are propmed.

INTRODUCTION The calculation of reserves in an oil reservoir or the determination of its performance and economics, requires a good knowledge of the fluids physical properties. Bubblepoint pressure, GOR, OFVF and compressibility are of primary importance in material balance calculation, whereas viscosity plays an important role in production
test interpretation and in properties are determined well problem analysis. Ideally. these from laboratory studies on samples

reliability only in a well-defined range of reservoir fluid characteristics. This is due to the fact that each correlation has been developed by using samples belonging to a restricted geographical area, with similar fluid compositions and API gravity. In particular for oils with gravity less than 22 API the literature is very poor and nearly absent for oils with gravity less than 10 API. This work is aimed at anaiysing tiie reliability of literature correlations, listed in table 1, relevant to heavy and extra-heavy Agips reservoir fluid samples, shown in table 2. This will make it prwible to evaluate the use of some correlations in -----;.. . . . :,.I. _C A 1 .,:.., *;-. 9.m ... UI nPI ~IWI1y l!! whl~ll IIU fi,.-,,li~,a,,,,. h .,. ha ~,,, tia.- -,, ~,+osed yet (except for viscosity): for oils with density lower than 10 API.

collected from the bottom of the wellbore or from the surface. Such experimental data am however not always available because of one or more of these reasons: a) sampies coiiecWi are not reiiabie. b) samples have not been taken because of cost saving. c) PVT analyses are not available when needed. This situation often occurs in production-test interpretation in exploration wells. In such cases PVT properties must be determined by using empirical derived correlations. Obviously the accuracy of such correlations is critical for the above mentioned calculations and it is not often known in advance. Despite the great number of work performed in the past 50 years on PVT correlations. each of them seems to be applicable with a good

raIIgCS

LITERATURE REVIEW The following


presents a review of the most known correlations published in literature. The range of input data used by each Author in developing his correlation are provided in tables 3 and 4.

References and illustrations at end of papaer

647

PRESSURE-VOLUME-TEMPERATURE

CORRELATIONS

FOR HEAVY

AND EXTRA

HEAVY

OILS

SPE 30316

published two correlations for determining, In 1947 Standi&my respectively, the bubblepoint pressure (Pb) and the oil-formation volume factor (OFVF) at bubblepoint, from known values of reservoir temperature (Tr), solution gas-oil ratio (GOR) at bubble point, oil gravity (w) and gas gravity OS). In ail. 105 ex~~nt~ly determined data points on 22 different cmde-oil/naturzl-gas mixtures from California were used. In 195S Laaate#a presented a new correlation for Pb. In all, ! 58 pressures tlom 137 measured bubblepoint experimentally independent crude oil systems from Canada, western and midcontinentrd U. S., and South America were used in his work. In 1959 Chew and ConttaU#5/ proposed a correlation to predict the gas-saturated oil viscosity (pol) as a function of dead-oil viscosity (,pod) and GOR. llte correlation was developed from 457 crude oil samples tkom Canada, USA and South America. fire study showed that at a fixed GOR, the relation between wol and the corresponding p od is a straight line on logarithmic co-ordhtates. In 1975 Begga and RobIrtao# published two new correlations for calculating pod and pol. The equations resulted tiurrt a study of 2533 viscosity measurements involving 600 different crude oil systems. An accuracy of -0.64% for the dead-oil viscosity correlation was found when tested against the data used for its work. When tested against 93 cases from literature, the average error increased to 114.27%. f?te Authors did not explain the reason for the large errors but simply warned that the extrapolation outside the range of the data mwd to develop the correlation should & done with care. In 1977 Vasqaez and BeggsY presented correlations for pre&cting GOR and OFVF of a gas-saturated crude oil, as a function of cmde oil API gravity, ?g, reservoir temperature and pressure (R). In total, 6004 data points were used. distributed into two groups (less than 30 API and greater than 30 API) because ofvariations in the volatility of crude oil. The Authors found w to be a strong correlating parameter in the development of the GOR correlation. Because w is dependent on the conditions under which the gas is separated from oil, a correlation to normalise ~ to a separation pressure of 114.7 psia was also developed by the Authors and tested against 124 data points from 27 different fluids. Vasquez and Beggs also investigated the viscosity (PO) and the isothermal compressibility (Co) of under saturated oils. using 4486 data points for the Co correlation and 3593 data points for the po correlation. In 19S0 Glttw@ presented correlations for estimating Pb, OFVF and pod, as a function of Tr, total surface gas gravity, GOR and API gravity. Because the first two correlations were developed using data from 45 oil samples with paraftinicities equivalent to North sea oils, an adjustment to the API gravity term WZ.. suggested for using the correlations with oils of a different compositional nature. Gkso alar) provided a method for correcting the predicted Pb for the presence of C02, N2 and H2S in the total surface gases. The correlation for pod was developed from data obtained from 26 crude oil samples. In 198S Egbogah and Jackml proposed two different correlations for estimating pod. fire first one was a nrorMied Beggs and Robinson correlation obtained by using 394 oil systems from laboratories of AGAT Engineering, Ltd. llre second one introduced a new parameter to estimate the Kod: the pour point temperature (Tp) which is, by definition, the lowest temperature at which the oil is observed to flow. Because Tp seemed to be related to crude oil paraffin content (it increases with the paraffin content), the Authors believed that important chemical compositional aspects of crude oil could be considered in the viscosity correlation by introducing this parameter. llw average error of the equation with Tp was slightly lower than the modified Beggs and Robinson correlation (-4.3% vs. -5. 13%). Since Tp is not an ez..ily-measurable parameter on field the latter correlation has not been investigated in this study. in 19SS Marhou#w published empirical correlations for estimating Pb, OFVF at bubblepoint and total OFVF for the Middle Bast crude oils, as a function of Tr. yg, GOR and API. A total of 69 PVT

analyses of bottomhole fluid samples were available for the development of congelations. Only the correlation for Pb has been considered in this work. In 1988 Asgarpour, McLauchlinj Wong and Cheum# presented a new set of correlations to estimate Pb, OFVF and GOR (at and below bubblepoint) as a function of yg, W, Tr and GOR. fire correlations were baaed on more than 310 different crude oil sampks from Western Canada. Because the physical properties of each geological formation in Western Canada exhibited different behaviottr, it was necessary to develop correlations for 3 different geological formations. Although the average errors of the correlations are very low, the paper has not been considered in this work since information about the geological formation of crude oil samples wem not available, and because this information is not easy to gain on fiPiIi . ..... In 1989 Labedi/lv published a new set of equations for estimating OFVF, oil density at and below bubblepoint, and Co of the African m.mtir ss ~ . ... . .. .. -. , .. . tkld @@ ss first.-, .,. fl.i;~ , - fimetinn of eYW~!v.nMM.RU~~e ,.. stage separator pressure and GOR, API, R and Tr. PVT data for 128 samples were collected from IJbya, N@eria and Angola reservoirs. Only the compressibility correlation has bum considered in this study. In 1990 Kartoatmodjt43 presented new empirical correlations for predicting OFVF, Pb, vod, wI, KO and Co as a timction of measurable parameters such as Tr, separator gas gravity (OOPsp), API and GOR. A total of about 1400 different samples were used to develop the correlations. Most of them were extmcted by PVT reports from South East Asia, California and Alaska and a reaaonabk group from literature. The new correlations were developed using the functional form of the previously published ones which gave the best estimate. Ilre Author also presented a correlation to convert OFVF and GOR from differential to flash liberation process at the separator condition. The OFVF, GOR and Pb correlations were developed using both flash vaporisation data and differential vaporisation data, (the latter converted to flash using the above mentioned conversion factor). Kartoatmodjo stated that these correlations are applicable to a flash process only. Applying these equations to a differential process might lead to errors of up to 20%. in 1990 Mti

Kattan and Salnm#4

proposed a new general


vol. GOR and experimentally sampb from correlation is

correlation for estimating po as a function of R, Pb, AP1. flu correlation was developed using 253 determined oil viscosity values on 41 different oil North Africa and Middle-East oil reservoirs. The

derived from plotting (R-Pb) Vs (po+tol) on a log-log paper. W plot shown a series of straight lines of a constant slope whose intercepts could bz represented as a function of APl and GOR.

McCaht Jr. and Creage#y developed an !n M)90 RoUi~ empirical equation to estimate stock-tank GOR as a limction of separator pressure and temperature {Psp, Tsp), API and GGPsp. llw correlation was obtained using a logarithmic model on a total of 301 black oil samples. The solution GOR, obtained by addhg the stocktznk GOR from equation to the field-determined separator GOR, has been affected by an average error of less than 3%.
In 1992 Labedi/l& published a new set of correlations to predict pod, pol and po. The data-bank for the development of correlations consisted of about one hundred laboratory analyses, representing the fluids of the entire producing reservoirs in Libya Each equation developed is a function of easily-obtainable datz such as API, R and Tr. In particular, with regard to the pol correlation, all equations previously published correlate @to pod and GOR. In this study ptrl is a direct fimction of pod, API and R, pararnetem more easilymeasurable in the field than GOR. Lzbed also published a relationship between differential and flash API. Even if the API used in all of the oil viscosity correlations developed in this study was obtained by flashing the fluid sample to the atmospheric pressure. which can be easily done in the field by flashing the well dirtxxly to the stock-tank, this relation makes it possible to utilise the viscosity

648

SPE30316

G. DE GHEITO,

F. PAONE, M. VILLA

dsta from the samples that are not flashed to the atmospheric pressure, but differentially liberated. The new correlations can be applied to other geographical areas such as the Middle East, the North Sea and some pats of North and South America, but they should be used within the limit of input ds~, in particular they should not be extrapolated for crudes of less than 32 APL In this study it was decided to extent the Labedis correlations to heavy and extra heavy oils. This was made because no literature correlations are available for oils with API < 14.4 (see tables 3 and 4), except for dead-oil viscosity (Egbogsh-Jack correlation). For this reason all the analysed correlations were applied over the range of input data reported by the Authors. in 1993 Petraaky asd Farshad/7 presented new empirical PVT correlations for estimating Pb, GOR, OFVF and Co, as a function of wd~hlI+=tn PVT analysis, m-l a.-- .. . .. G-IA ---- A h-d Of ~ I !~bo~tory rvm,,l-ll, ---made on crude oils extracted from reservoirs offshote Texas and Louisiana were used to deveiop tiie correlations. Audhors fount+ tiiai thdr correlations could predict the PVT properties with average absolute errors ranging tlom 0.64% for OFVF to 6.66% for Co. The correlations were developed specifically for Gulf of Mexico crude oils but Authors said that the same equations could be used in other regions of the world. Od y the compressibility correlation has been considered in this work.

. .

in the past few years, oil companies have become increasingly interested in reservoirs with the extra-heavy oilsnsn. there am no correlations in literature which cover the range of oils with API S 10, except for viscosity.

The reliability of each correlation and for each parameter was therefore tited for each API gravity class. No analyses were made for the whole group because it is plausible that samples belonging to the asme class are physically and chemically more comparable than samples tlom dtfferent classes. l%e reliability study was carried out using graphic and statistical instruments. Calculated (~) vs. meawm%l (Mi) - value diagrams wete created for each pmmeter studied in order to have a clear and immediate view of the behaviour of each correlation. For reasons of space, not all the calculated-value vs. measured-value graphs, relative I..* a,.1 ;, w . I- ..- G.. :...AA.AA :. qI.,:. - III W. ptp,. .,,.t~, ,. _& to c%chCOfRi~tiWi, nave ucm mmmwu to show a single diagram which gathers the beat reds decided obtained for in&idual classes of oil . The diagrams for each proprxty estimated are shown in figures 1,3,5,7,9, 11, 13 and 14. The qualitative analysis carried out by means of diagrams was accompanied by a statistical analysis, of which the starting Pint was the dative deviation between estimated and experimental value (Ei), thus defined (1)

IO!LIABILITV ANALYSIS LITERATURE m cortstmmorm


Ttds work analyses the most well-known correlations described in literature for estimating PVT properties such as bubblepoint pressure, oil formation volume factor and solution gas-oil ratio at bubblepoint, dead-oil viscosity, gas-saturated oil viscosity, under saturated oil viscosity and isothermal compressibility. It does not however include those correlations which require. as input dst& psrametem which ate not easily measurable on field or not obtsinsbie from PVT reports. Table 1 shows schematically the Authors and the relative correlations considered for each property examined. Starting exclusively with the PVT studies carried out over the last 30 years on Agip oils, a selection was made excluding those lacking all the input data necessary to use PVT correlations. In this way, a very heterogeneous sample of 63 crude oils was set up, representative of diverse reservoir conditions, in order to ensure that the conclusions obtained from this analysis would be generally valid and have an extensive applicability to wide range of operative situations. The 63 oils come from the Meditemanean Basin, Africa and the Persian Gulf. Table 2 lists the range of input and output pamneters of -*! AJ.-1. -:1 --.-1 . ..,I.:1- .r..l.la < . A. -.+tall ,,,wmw,,,rn,,w.y= all Aslp S Ull 511111@% WIIIIG I autG J ,+.,. measured PVT data involved in the present study (about 1200 data pohta). Tables 3 and 4 list the range of input and output parameters upon which each Author based the development of his correlation (Authors defined range). ..__,... _* __ _,, ,- - &.-#-_--., I ne aemmy or an UII IS u IUIWIHIUI CMSCCG*AC its R dktt its chemical composition, on which all the fluids main properties depend. For this reason, the API gravity was chosen in this study among all the different parameters used for classifying oil% tberefom Agips oil sample was divided into 2 different classes of API gravity as follows
q extra-heavy oils API s 10 10< s heavy dla API ~ 22.3 l%e second class correspond to a standard classification of Oil/-won the basis of the API gravi~, the extremes of the ranges which identify the class can vary as there is no univerwdly recognised classification. Even if the class of extra heavy oils does not compare in the standard classifications, in this study it was decided to mal yse separately oils with API < 10 mainly for the following reason5

After having calculated the Ei for all the available samples, results wtne subjected to a statistics] analysis calculating the average arithmetical value (Q of the I+ and their standard deviation (SD), i.e., the dispersion of the ~ around their average value ~, using the following equations (2)

SD= {

~i~,[Ei

%]

N-1

(3)

l%e correlation providing the smallest ~ value was judged to be the best. When equal ~ was found for more correlations, the lowest standard deviation value defined the best one. Table 6 provides the best results obtdned from the statistical analysis, for the different parameters estimated, for the two API gravity classes. Below is a discussion of the rtaulta obtained for each property estimated.

kWLTS

OF RELIABILITY ANALYSIS PSIWORMED ON AGIPS AMPLES S

All the results are discussed with reference to Table 6 and to figures 1,3,5,7,9, 11,13 snd14c

Bnbblepointpressrm
Stsndhrgs correlatiodlw has given the best results with average errors of 9.1% for extra-heavy oils and 15.1% for heavy oils.

Sofstdan gas-oil ratio ~ best results are provided by the Standing and Vasquez-Beggs correlations with errors of 13.7% for extra-heavy oils and 25.7% for heavy oils. OUfmttation vohtme factor at bubblepoint Of the seven propenies anslysed, tlds one was estimated in the best
way. The highest errors did not exceed 1.5%. Vasquez-Beggss yu~#~ f?ve -m of tin.? than half of those indicated by the

Isothermal C-Sdbif@
IRe estimation errors nmge from 25.5% for heavy oils to 38.7 for extra-heavy oils. Vasqrtez-Beggss correlation gave the best performance for the both classes. mad-d Viseos@ The estimation of this property exhibited the highest ewor, the Ioweat errors bektg greater than 30% . Ilre errors are very high, especially

variations in the properties of crudes

preame of the most heavy hydrocarbonst2%%%i9iTfly n k

649

PRESSURE-VOLUME-TEMPERATURE

CORRELATIONS

~R

HEAVY

AND EXTRA

HEAVY

OILS

SPE 30316

with regard to the cla..s of heavy oils. This behaviour is justifiable bearing in mind that the correlations estimate this property with only two input variables: API and reservoir temperature. f?re correct measurement of this property is difficult to achieve even in the laboratory.

estimate the same property, is certainly more reliable than a single correlation for all the sample. For this reason, new equations were pruposed only for each API gravity class and not for all the group of Agips oils. In order to test the reliability of the modified quations, the same graphic-statistical instruments as those in the previous study were used. The results obtained are shown in Table 7 and in fig. 2,4,6, 8, 10 attd 12, prepared in the same way as those for the analysis on the literature equations, in order to be able to compare the two sets of graphs more adequately. In some cases, it was necemry to eliminate ml. bin. n..l$&~ ~~ o*~ @ rn&~ !a~ .. Uwn,yrd . . . . ...1. #.-n- *h.= -s..s.. . -,.,~ W,,lc ,,,,, .,...., regression more reliabl~ however, the exclusions never exceeded 5% of the entire group. The study did not take into consideration the correlations which estimate the oil formation volume factor at bubble point as the estimation of this prrpwty carried out using the equations chosen from literature was felt to be very satisfactory. Appendix A shows the analytical form of the new correlations.

Gas-safrrmted oil viscosity The average errors of the best correlations range between 14% attd
16%. The best results were provided by Kartoatmodjos correlations, :v~~rs comparable with those found by the Author in his own Figure 9 shows the distribution of the @nts calculated with tire best correlations where the input variables (dead-oil viscosity and solution gas-oil ratio) are measured values obtained from PVT repnrts. Figure 14 shows the results of the same correlations where the calculated value was used as input data of the dead-oil viscosity. Noteworthy is the increase in dispersions of the points around the bisector which corresponds to an average error increase of more than 15 percentage points. The difference is due to the fact that by incltrdlng a calculated rather than a measured input in an equation, the estimation error of the equation in some way combines with that made on the calculated input even if the latter has been calculated with the best correlation. lle greater the error on this input, the greater the correlation error. Since the correlations which estimate the viscosity values at different presstrms are all inter-connected, the lower the estimation error of the dead-oil viscosity, the better the estimation of the gas-saturated oil viscosity. The same applies to the correlations relative to the undersatttrated oil viscosity which have the gas-saturated oil viscosity among the inputs. This proves the importance of correctly determining the dead-oil viscosity, which on the other hand, is the property calculated in the worst way. The observations made can be naturally and easily extended to all the other Ptopertie% in fact. a quantity estimated by using measured input variables will undoubted y be more reliable than one estimated with ~~i~~l~fd inputs.

~LTS OF RIZLIABILITY ANALYSJSPERFORMED ON MODJFIED CORRELAITONS.


results of Tab. 7 obtained for the different properties are shown llre below, and are compared with those of Table 6.

Bubblepoint pressure fire starting models used for improving the estimate of this property
was Standings correlations for the both classes of oils. Tire new correlations twduced the estimation errors of 4.9 percentage points (aes Tab. 6 and 7) for the class of heavy oils. Regression in the class of extra-heavy oils, having given results worse than the starting c~si~ model, is not shown. Standings correlation WM sufficiently reliable for atimating oils bubblepoint pressure with API c 10. Comparing the diagrams in fig. 1 and 2 it can be seen that the most significant improvement in the new correlation is in the pressure range below 20(N paia. In order to allow an easy interpretation of the results obtained with the reiiabilit y studies perfomred in this work, the best results of the statistical analyses are compared in a histogram for each PVT

Undkrsatnmted oil viscosfty The best correlations showed a maximum error of 12.3% (Labe&
extm-heavy oils). Note that Labedis correlationllw which had in fact been gauged with oils with API >32 (Tab. 4), showed excellent results even for the other classes of oil. It should also be pointed out that the ertur in estimating the viscosity normally becomes smaller attd smaller as we go from atmospheric pressure viscosity to reservoir pressure viscosity. It is likely that the input variables which estimate the reservoir oil viscosity (bubble point pressure. reservoir pressure attd 00R), characterise the phenomenon better than the inputs of the dead-oil viscosity (API attd reservoir temperature).

Prowy (= fig.15 to 20). Each histogram shows the value of the most important statistical parameter (~. average absolute error) for the two classes of oil into which the sample was divided. Sofulibn gus-ail #a flte quations used as model were those of Standing for extra-heavy oils and Vasqtrez-Beggs for heavy oils. llre regression of VasquezBeggs qttation was carried out keeping fixed the quation of the
~orr provided by the Authors; this was done every time the starting model was a Vaaquez-Beggs correlation. llre new quations reduced the estimation error from a minimum of 7.2 to a maximum of 8.7 percentage points. The comparison between the diagrams in iig.3 and 4 shows that the most obvious improvements wete in the 00R range bdOW 250 scf/STB.

~VELOPMENT OF MODIFIED CORRELATIONS


fhe results obtained from the atmve-explained reliability analysis shows that, except for the OFVF comelation, the average errors in determining PVT properties are still high, especially when oils are beyond the Authors defined range. For this reason the need to improve the reliability of the literature correlations has been y recognised. The functional forms of the correlations that in the previous reliability analysis on Agips samples gave the best results, for each PVT property, have heen used as models for a best-fit activity aimed at improving the accuracy of literature correlations in predcting PVT properties for typical Agips oils. Maintaining the same functional pattern of the starling model, the numerical Ctrefflcients of the different quations were m-calculated by applying multiple. linear and non-linear regressions by means of the SAS program which carries out these regression analyses using the minimum squared method. The modified correlations were obtained for each class of density into which the Agips oil sample was divided. In fact oils from the same class are more comparable than oils from dKferent classes, and then the availability of two different equations, one for each class, to

isothermal eonrpressibility Ilre model to regress was Vasquez-Beggs correlation for both the classes of oils. IIds setof new quations provided the most
significant improvements. The error decreased from a minimum of 10 to a maximum of 30 percentage points for extra-heavy oils, Comparing the diagrams in fig.5 and 6 it can seen that the greatest :- .----* - ... Mid %sm.-l llllPIU+GIIlclIcm = OU.=lJW %. -vmY. worn. lit WW-n =, ,V, M -a.lh vs,..J -. .. W.. - ..,,., i Q ~ 10-6 paia-t.

Dead-d viscosfty The models chosen was Egbogah-Jacks correlation for the both
classes. The dead-oil viscosity is the most critical property to estimate with empirical equations. In fact, although the errors dropped down to 13 pementage points with the new quatiorrs (extra heavy oils), values higher than 30% (heavy oils), me still present. On the other hand, the viscosity, not being a state property also depends on the behaviour of the fluid. All the correlations assume that the fluid can be considered Newtonian, but this is not always true, especially where high viscosity are concerned. To attempt to estimate a quantity

650

SPE 30316

G. DE GHEITG,

F. PAONE, M. VILLA

of this kind using equations which only use two input variables (API and reservoir temperature) becomes even more difficult. In any case. not even laboratory measurements of viscosity can be considered completely reliable: in fact. particularly in the range of high viscosity, differences of 10% between two measurements taken on the same sample by two different quipment, gauged in the same way, are normal. The diagrams in figures 7 and 8 compare the trend between the old and the new quations. They reveal that the most significant improvements are to be found in the range of viscosity greater than Iocp.

The reliability analysis of the literature PVT correlations earned out on 63 oil samples from Mediterranean Basin, Africa and Persian Gulf, gave the best results for the estimate of the OFVF. with maximum errors lower than 1.5%. The estimates of Pb, pol and VO exhibited maximum errors of about 15%, 16% and 12% respective y. The GOR, Co and pod estimates were less precise the maximum errors were about 26%, 39% and 42% respectively. Ilse new PVT correlations proposed in the paper gave errors lower, on average, than 10 percentage points when compared with the best literature correlation for each PVT propefiy. In particular. for the isothermal compressibility of extra-heavy oils. the new correlation revealed an error lower than 30 percentage points. It is believed that the new correlations are sutlkiently extendible as they were obtained on a very heterogeneous sample of oils. A deep literature review has shown that, except for viscosity, there are no PVT correlations for extra-heavy oils (API S 10). The proposed new equations for such oils provide average error of 6.5% for solution GOR, 8.5% for isothermrd compressibility, 17.4% for dead-oil viscosity, 12.6% for gas-saturated oil viscosity and 4% for undersaturated oil viscosity. A further investigation of the new modified correlations, performed on a new different group of oil samples (ftum literature and Agips reports), has shown that the results obtained with the new equations have a general validity. lWs analysis involved only the viscosity correlation because of lack of literature data about the estimation of the others PVT properties.

Gsw-sotrmrted oil viscosity The starting model for the regression was Kartoatmodjos correlation
for the both classes. For the Kartoatmodjos correlation. the multiple non-linear regression was carried out by keeping the quation supplied by the Author fixed for the input variable ~corr. This procedure was also followed for the other properties whenever the starting model was one of Kartostrnmljos equations. me regression reduced the estimation error from a minimum of 2.1 (extra heavy oils) to a maximum of 4.3 (heavy oils) percentage Pints (see Tables 6 and 7). D~agrams in fig. 9 and 10 show that the new correlations improve the estimate in the range between 10 and 100 cp.

Understslnmted oil viscosi@ Ilre models to regress were Labedis correlation for extra heavy oils
and Kssrtoatmodjos correlation for heavy oils. Ilre new quations brought the maximum estimation error to 6% (Tab. 7). The diagrams in fig. I I and 12, which compare the trend of the old and new quations, show that the improvements are distributed along the entire viscosity range.

FURTNER INVESTIGATION TNE NEW MOBWIED CORRELATION ON TNAT ESTTMATETHE VISCOSSTV The new modified correlations have been obtdned analysing Agips
oils sample. For a snore general validity of the results obtained in the previous analysis. it was decided to test the new quations using a sww group of oi Is collected tlom literature. A deep literature review has shown that the Authors are usually reluctant to publish the oil data bank used for testing their correlations. For this reason it was possible to collect thm Iiteratum only 10 oil samples. with dsta available for the ordy viscosity correlation analysis. To make more representative the results of this analysis. a group of 45 oils samples, collected from the Agips viscosity meawrements reports, has been added to the oils from literature. In this way an heterogeneous sample of 55 oils has been obtained. Ilse complete data bank is given in Table 8. Since the extra heavy oils are only 5, results obtained in this representatives as those of the class have to be considered not as heavy oils class (50 samples). The results of the statistical analysis, performed on this sample using the same statistical index as before, am given in Table 9. Comparing this results with those listed in Table 7 and, secondly, Table 6, we can say that:
q

NOMENCLATURE
API a co Ei Em, AAE GOR, LQ8 Ln Mi N OFVF, Bo, Pb Pr, P Psp Rst Rsp S.D Tr, T Tp Tsp YC02 YH2S YN2
Botb ft$,

Stock-tank oil gravity, API Calculated value Isothermal compressibility of undersatursted oil, psia- 1 Relative deviation between experimental value, % Average absolute error, % Rtot Solution gas-oil ratio from flash test, scf/STB. Logarithm on base 10 Natural logarithm Experimental value Number of data points Bubblepoint bbllsm oil formation volume factor. estimated and

Bubblepoint prwssure,psia. Resewoir pressure, psia. Separator pressure, psia. Stock-tank gas-oil ratio, scf/STB. Separator gas-oil ratio, scf/STB. Standard deviation Reservoir temperature, F. Poor point temperature, F Separator temperature, F. Mole tkaction of C02 in total surface gases, % mol: Glasoslw bubblepoint correlation. Mole tiadon of H2S in total surface gases, % mol: GlasosW bubblepoint correlation. Mole fraction of N2 in total surface gases, % mol: Glaso~ bubblepoint correlation Average
S@fiC gSWity Of total

Dead dl viscosity : the Em increased by about 9 percentage points for extra heavy oils and decreased by 2.4 points for heavy oils. l%e result for the heavy oils is very good and confirm the general validity of the new corresponding correlation. For the extra heavy oils the poor number of samples makes the results less representatives. However results by Table 9 for extra heavy oils are better than the corresponding by Table 6, relatives to the best literature correlations. Sdssrded dl viacoaity : Em increased by 7.2 percentage points for extra heavy oils and by 8.7 point.. for heavy oils.

Usdesmtssratsd

dl viscosity : an increase of 1.9 percentage

points for the Em in the clRssof extra heavy and a &crease of 0.3 points in the class of heavy oils confirm the general validity of the new corresponding correlations.

~. GG(av)
ygcorr, GGcorr

SU&2C @eS.

Gas Specific gravity at separator pressure of


114.7 psia.

651

PRESSURE-VOLUME-TEMPERATURE

CORRELATIONS

FOR HEAVY

AND EXTRA

HEAVY

OILS

SPE 30316

n~p, -fo, jtxt

GG(Psp),wp

G= Specific grWity Zi amy *~ZiiiGi ~~~~.@q~ Q~!Spific gravity. Undersaturatcd oil viscosity, cp.

! c L~H_~ pressiim.

R,

lm~ved

correlations for predicting the viscosity of

light crudes, Journal of Petroleum Science and Engineering, 8 (i992k pp 22i -234. 17 Petrosky G.E. Jr.. Farshad F.F.: Pressrrre-Volume-Tempemtum Correlations for Gulf of Mexico Crude oils, SPE 26644, (1993), pp 395-406. 18 Beat C.: The Viscosity of Air, Water, Natural Gas, Crude Oil and Its Associated Gases at 011 Field Temperature and pressures; Oil and Gas Property Evaluation and Resewe Estimates, Reprint Series, SPE, Richardson, TX, (1970). 19 Slotte in Frick T.C.: petroleum production Handbook SPEAIME, ( 1%2), Vol 2. 20 Calhoun J.C. JC Fundamental of Reservoir Engineering: University of Oklahoma Press, Norman, OK (1947) 35. 21 Trube A. S.: Compressibility of Under saturated Hydrocarbon Reservoir Fluids, Transaction AIME (1957) 210, pp 341-44. 22 Majeed G.H.A. & Salman N. H.: An empirical Correlation for Oil FVF Prediction, Journal of Petroleum Technology.

po, Vo
pod, Vod pol, Vol

Dead-oil or gas-free oil viscosity, cp. Gas-saturated oil viscosity, cp.

S1 METRIC CONVERSION FACIORS

Nm3/m3
q q q

(*)=gcm3
x 5.5519=
SCf/STB

KPa x O.14504= psia psia - 14.7= psig C XI.8+32=F KPa -1 x 6.894757= psia -1

q q

cpxl.O=mPaxs. bbl x 0.1589873= m3

23 Obomanu

D.A. & Okpobori G. A.: Correlating the PVT properties of Nigerian CrudesU Transaction ASME (1987) Vol lo9, pp214-16.

REFERENCES 1 Standing M. B.: Volumetric and Phase Behaviour of Oil Field


Hydrocarbon System, SPE-AIME, Ninth Printing (1981). Standing M. B,: Oil-System Correlations Petroleum Pmd~#ction Handbook, Frick T.C.(ed.), SPE, Richardson, TX (1%2) Vol. 2, Cap 19. 3 Standing M. B.: A Pressure-Voirrme-Temperature Comeiation for Mixtures of Crdifomia Oils and Gases, Drill& Prod. Pmct., API (1947), pp 275-87. 4 Laaater J.A.: Bubble Point Pressure Correlation, Transaction AIME (1958) 213, pp 379-81.

24 Ali J.K.: Evaluation of Correlation for Estimating the Viscosity of Hydrocarbon Fluids: Journal of Petroleum- Science and Engineering, 5 (1991),pp351-69. 25 Sutton R.P, and Farahad F.: Evaluation of Empirically Derived PVT Pmoetties for Gulf of Mexico Crude Oils, SPE Reservoir En~nee~ng, (February 1990), pp 79-86. 26 Callegari A., De Ghetto G.:Studio di Affidabilit~ di Correlazioni per la Stima delle Pmprieth di Oli di Giacinrento, Agip (internal report), (Gennaio I 992). 27 Lang K.R., Donohue D.A.T., P.H.D., J.D., Senior Series Edito~PE 406-Petroleum Engineering IHRDC E and P Video Library edizione in Lingua Italiana a curs di G.Flarnmengo (LACH) e ADFO.M.R. 28 Davis J.C.:Statistics and Data Analysis in Geology, John Wiley & Sons, New York (1973), IJP54-127 29 Spiegel: Statistics, Collana kchaum, (May 1976). 30 Chierici G.L., Ciucci G. M., Sclocchi G.: Two-Phase Vertical Flow in Oils Wells-Mlction of Pressure Drop, JVT (August 1974), pp 927-38, Transaction AIME, 257. 31 Chierici G.L.:Principi di Ingegneria dd Giacirnenti Petroliferi, Vol 1, AgipS.P.A, (aettembrz 1991 ). 32 Paone F.: Studio di Affidabilitk delle Correlazioni che Stimano Ie Pmprietk degli Oli di Gktcimento, Tesi di Laurea in Ingegneria Mineraria, Univeraiti degli Studi di Bologn& (13 ottobre 1993). 33 Closmann P.J., Seba R.D.: A correlation of viscosity and molecular weight, he Journal of Canadkm Petroleum Technology, (July-August 1990), VOi. 29, No. 4, PP 115-116. 34 McCain W.D. Jr., Reservoir-fluid property correlations-State of t~ Afi, S~ Rmoir Engineering, (May 1991). Op 266-272. 35 Puttagunta V. R., Miadonye A., B. Singh : Simple concept predicts viscosity of heavy Oii and bitumen, Oii & Gas Joumai (Mar . 1. 1993) no 71-73. . ... . . . . ..- ,3r r.36 A1-Blehed Viscosity petroleum 37
1<7 -..,-. 16? .-,

5 Chew J. & Connally C. A.:A Viscosity Correlation for GasSaturated Cmde Oils Transactions AIME, (1959) Vol. 216, pp 23-25. 6 Begga H.D. & Robinson J.R.:Estimating the Viscosity of Crude Oil Systems JPT, (September 1975), pp 1140-41. Property Prediction, SPE 6719, (1977). Pressrrre-Volrrme-Temperature Generalised 8 Glaao O.: Correlations JPT (May 1980), pp 785-95.

7 Vaaquez M.E. & Beggs H. D.:Comelations for Fluid Physical

9 Egbogah E.O. & Jack T.Ng An Improved Tempemtrrre-Viscosity Correlation for Crude Oil Systems, Journal of Petroleum Science and Fxgineering, 5 (1990). pp 197-200. 10 A1-Marhorrn MA.: PVT Correlations for Middle East Cmde
oils? JPT (May 1988), pp 650-66. 11 Asgapur S., McLaughlin L., Wong D., Cheung V.: PressrrreVoiume-Tem~mm Correlations for Western Canadian GSSCS and oils Petroleum Sochsty of CIM, paper No 88-39-62 (1988), pp ~~- i i~~-~& I z W! R.: use of Production Data to Estimate Volume Factor, of Reservoir Fluids, Journal of Petroleum Science and Engineering, 4 (1990), pp 375-90. Density and Compressibility 13 Kattoatrnodjo T. New Correlations for Estimating Hydrocarbon Liquid Pmperdes (71eaki), Ilte University of Tul~ The Graduate School. (1990) !A Maj~ G.B,A,: Kattan R.R. and Sahnan N.H.: New correlation for estimating the viscosity of under saturated crude oils, Journal of Canadian Pettolerun Technology, (May-June 1990), Vol 29. No.3, pp 80-85. 15 Rollins J.B., McCtdn W.D.Jr., Creeger J.T.: Estimation of

singhB.,

M.S., Sayyouh M. H., Deaorrky S.M.: API Gravity and Deterndne Cmde Oil Sulphur Concentration. Engineer International, (June 1993), pp 5660. Mittdmrye A., Puttagunta V.R.: Heavy Oil ViSCOSity

range from one teat, Hydrrwarbon processing, (August 1993), pp 38 McCain W.D. Jr.: Chendcal Composition Ddmmkea Behaviour of Reservoir Fluids, Petroleum Engineer International, (October l@lm nn l&25. ..=,, rr -- --39 McCain W.D. Jr.: Black Oils and Volatile Oils-Whats the mffercnce? Petroleum Engineer International, (November 1993), Pp 24-27.

80hltiOll of Black OilS, JPT (Jfmu?uY GOR 1990), w 92-94.

SPE30316

G.DE

GHETTG,F.P AONE.M.VILLA

40 McCain W.D. Jr., Bridges B.: Volatile oils and Retrograde Gases-Whats the Diffemnce7 Petroleum Engineer International. (January i994j. pp 35-36.

F+0.001567. F2 Vol = 2.3945+0.8927.


Willxe

(A -8)

----- , F= -0.0335+1.0785.10-0-% .py~~~~+ ) ( ) R.! -0.00081.


y=lo AP1w8 .(7)-02*.bg * H]

APPENDIX A - MODIFtEDCORRELATIONS l-Bubbiepaint Pressure:


q

~gcorr = TRP.rp ~ 1+0.i595 [


q

Heavy oikx

Pb= 15.7286. K

1[)]
,00.0020T
,00.0142.API Yg .oiw.An-o.m156. )

Modified Stanclktgs correlation 0.7885


(A-1)

Heavy dk

Modified Kartoatmodjos correlation


F2 (A -9)

Pol = -0.6311 +1.078 .F-0.003653. Whete F= ( 0.2478 +0.611410-0 000845m -0.W08 i ~Rs r

.p~+0558y)

y=io

2- solution GOR.
q

~xcorr = ~gP.sp

i+0.i595 APIOm8 .(Tv)-02%mg

ExtrR-hezvy oik Modified Stsndings correlation


T)) 2s (#$ . ~) 64Jnderaxturated

(~]~ ii4.7

( Pb ,.(O . s= g [ 10.7025
q

oil viscosity:

Extra-hexvy oikx Modified Labedis correlation

Rs=%=+
where

Hezvy oils: Modifti Vsaquez-Beggs correlation 1.2057 ,,0 lo.9267. AFl/(T+460)


56.434

(A -3) . .

Po=Poi

[[

J-;

1[
io-2. i9 P [

i.055. m0.3132

pod io0.0099. AFt

v A yRCoW = YKp~p 1+0.5912.PI .Tfp LOX

[11
114.7
-81.4476. API

,.-4

Heavy oils: Modified Jktoatm@os

corrziation

II

(A - 10)

P. = 0.988ti. Po, +0.002763 .( P- Fb) . -0. Oi i53v~3

+0.03 i6qI~i5939 )

(A-ii)

3-Iaothermd Compressibility:
q

Extm-heavy oikx Modified Vaaquez-Begga correlation


889.6+3.1374.
%+

~~corrYRP,fp =
(A -4)

. I +0, 1595, Ap, 0.4078

20Tg Pg

-627.3 105

.ygw=

co=

(~fp)-02% * 11
[ ii4.7

where YgP.Tp [
q

Y~corr

1+0.5912 APl. Tv. f.oR

Heavy oils: Modifkd Vssquez-lkggs


-2M1.8+2.MR s+25.W39Tg-1230.5yg -

T (11
114.7 ,.-4 correlation
+41.91 API

co.

(A -5)

Pg.los where fP ~Rcorr = 7RP.VJ . 1+0.5912 .APIF~p. Log 114,7 [11 [ ,0-4

Mkad-oil
q

viscosity

Extra-heavy ok
+1]= (A- 6)

Modiikcl Egbogah-Jazks correlation


<...0,--1 ,4P; -0,01 I*B. IW\Tgf )

Iwlw(Pd
q

i,9ij296_- U.UIXM9 -----

Heavy dim Modiiied Egbogah-Jacks corteiation


Pd () + i = 2.06492-0.0179 API -0.70226 .iog(Tg) (A- 7)

tog iog

S-Gas-sxtorxted dl 4kOZity
. Extra-heaw . oils: Modified Kartoxtmodios correlation .

653

ITABIZ 1: FUJIDPROPERIY CORRELATIONS

lpbdd roper& P
IBubblepoint
premlrc

1,Las2tcr~, w20@ Staaiiiiiziii


Ibtodmdjo

lwJU-Marhounnw
,
n~

SolutionGOR

s-

V--

Km@modjo, Rollinf5-MabCm&r IOFVF

SVqudkggs, htoatmdjo
Vuqucz-13qEW

Glaso,

Kart@mdjo,

cmqm?siity

Lsbali z,

Fetros&Fsrsha@7

un&@Ulmdoil
Vi2c08ity

Vasqued3qWSKmt@modjo, Majccd-Kattan-Salman 14, I.abali fl~

ITAB152

AOIPS ANGE POR PVT PROPERTIES R SAMPI.X

I
I I

r-ati) Lwlirtalmlumm

] 1038.49 t07411.34 I 131.4ta 230.7

w
59 to 177.8 11.1 to 575.62 4.39 t031L41 0.5 to 9RS 7.7to 13s6.9 21tow5.9

I I i

Mobhctkmafco2rntdxl

gaam(%mal.)

I
I I

oiifmlldkmvolIlmosdar@bl/sTB)

11.057to 1.362 10 % I 3.02 to 429

%iliMu&x .Dedd ViIcody ((p)


~clilvimaily(cp) Udmdmidosviaccdy

((al)

12.4ti334.6

TABLE3 AUTHORS DEFINED WOE

FOR BUBBLEKXNT PRELSm -

X.LITION OOR,OFVFAND COMPRESSLBIIIIY CORRELATIONS


O&o
22.3 to 48.1 165t07142 8oto280 1.025 to 2.588 90 to 2637

117.9to 51.1 48t05780 82t0272 I13t02905

K@@-#o
14.4 to 58.95 oto6040 75 to 320 1.022 to 2.747 oto2890 0.4824to L668

vMqurz-Beg@ 15.3 to 59.5 15t06055 170 (Oll!aa) 1.028 to 2.226 oto2199 o.511to L351 . 6oto565 76 to 150 141t09515

AIMafhmlll 19.4 to 44.6 130 to 3!573 74to 240 1.032 to 1.997 26 to Mm 0.752 to 1.367 20t03573

Rolsm-MdhilI lsto53.5 0.579to 1.124 29.7to 3114.7 6oto ml 4t0220

-F16.3 to 45 1574to 6523 l14to 288 1.l178to 1.6229 217to 1406 0.5781 to 0.8519 -

~~
32.2 to 48 520106358 12810306 1.088 to 2.92 -

Tank-d ~vity ~APl) BubblepOidpmsum (pllia) ~~m OFVF d

16.5 to 63.8 130t07000 100t0258

Ilubbw (bwsTB)

I 1.024 t02.15 120to 1425

Solution OOR(dSTB)

0.65 to 1.276 415 (ImaO) 125 (mean) .

100 3s to 294 loto6000 .

34.7 to 789.7 60 tal220

1700to 10692

Sto&tmk OOR (sc@STB) Sqmmtor OOR(IC@TB)

12 to 1742

Solution 00R(acflSTB)
BIMM ~ (W@

De40s ViaxJdy (q) ouutw@60sviwodh(cP)

TABLE !k EXPERIMENTALLY NlEASURED PVT DATA

4mdl

1 1
414M4 2228, 4484.78

-q

4823,23

m m.m 52.22 102.1O 84.04


38,27

2mm

14,1

2o121 6, m. e

482642 48W2 ma
4882M a23,81

8L22 0,27
-.33

24.48 L224 Ia40 2.4m U44 L4m

22L4 22L0 Im,o 2a2 22L0 v7,2 14L2 =0 m.8 862 84.0 12L2 m 2ao Imo M2 2ao 3s.0 228.0 1248 424 IM3 1%0 Xl 1s0 247.0 MO W1.o me 3.%6 m.8 ll&6 MO 3Z0 10Lo 4L0 104, 4 la 4 u o * o 4%8 48.8 20,4 m.8 U4.4 3L2 m o 72.2 n 6 20.6 100. 4 73.4 e .2 M, o a o la 4 8Lo 28,2 24.o!

4a04 W# 23.02mm 82.27 34W


20,83 am 22.37 7L07 280222 6W,22 848,42 723.20

442
7,44 24.42 Lm 44.1I %32 860 1 82.44 83.S m.m W53 4640 32.09 M 7 w am e.n 052 0.41 10,% O@ 12.3I 20.04 3Ln 4,924 LL LL 3,m4

28un

law 4nm 240.00


22L24

200.

448

7,77 27.44

mm

224.14 23L47
334,21 24L80 mu m%m mt,lb 17.21 lm.m 43.87 4U3 sot 2L21 *U 2L48 10L82 228.00 $732 220.24 M4,40 m,la llLfi mm Mm

ann

2?2. = m m.u

LI* L2X 22.37 M12 4.828 w 28.21 0.722 12.02 L2m %33 4.322 wm am

W.a no7 7L87 w 287.24 W7 424.22 3$37 847,42 2L02 llw,n Isa 224.44 lam,m 224.44 320217
227.71 24.s7 22L44 124,42 M18,84 m14,72 224L80 142L22 Ilmw

mb7!

480SM

4329.4! 4410.6

-X 228t.a
2727.2 2m2

1 I

4.22

0.734 L2m

12.1

4,11

% 3 w

24.24 La3 lqn Ll~


7,44 22.81 28.48 4.0s 44.42 ma 2aa W 31.42 20.43 M 48041 23X4 an W24 32.80 2M2 22,22 mu 14,44 m 3L42 2L22 l,m 4234 84.42 142,23 L3m w L044 L2m 0288 @m L128 2.217 L281 * Lm L2m 8.824 L4m 3,422 3.41 1 L828 2.41 7 L143 L202 Lom 2323 3347 L222 Lm2 w

3142.O

424

4
2n8.6 741M

122L0

m.m 4a04 :78.22 mL42 227,7i 3S7 lm4,22 8sS7 U8M3 224,44 23?S4 U.21 2n&m 20.X Z7.7 1 w 2m,o1 3a.08 42.22 222.22 m 1724.2 1 28am 2m%m m.n 4m.44 1074.72 m la8Loo 2S37 224.84 M&a M644 888,32 1047,1 9

ua
=4 1 K2 1 2292 Ua la 22 0.02 M7 8,n 88.22 S.23 41 7 28.n 3232 4s.m 2.92 mn 2oYl 8L23 4a23 %w #1#4 1L20 8.20 73.47 47.01 %24 44.24 Z20 4.48 w

7413.2

all

m47.4

4227.2 17822 mm.! 1448,!

22M3 244.23. 1 1

288.

97,1

72.07 Mm mm 73.07
lLOl 7Lol 227,n w 22L41 m m, nm3

03?.8 1 nw.28 784.27 488,s3


784.27 ms.m 1034,43 784.27 12a w % n ,42

!!liiI
Ll a a 4.L L81 t L L L L L1 L L

m
1

m2u

24
1

mm!

4m7.ml mm 72274
23801 1748/

4824,(M

44?234 2mu
2208J

W.28 24.42 27.84 mm 202m 22La m


I

4%

1L44

n 07847 .42 1% u 122L04 94.22 nw w 823. m m, n M63 N 224.44 - .m 2423.22 ?a30 3333 1 w
464 1 Mm ,48 483 1424. 33.


TABLE 6: BEST RESULTS OF STATISTICAL ANAUSYS FERllXIMED ON AOIFSSAMPUN
(AAE =Aq Om <-10 API AAE SD Aulha SD T . 9L 15.1 13.9 tb6-OilVkOdly <-10 AM ygr SD pgc 10 c OAFIC= 22.3 SD 24:9 : + Abduta k. % SD - *W ~@ OFVF ~ 1.2 vuqufA&y 1.4 1.1 Wdm@mddosLabus 12.3 7.8 10.1 105 I&4k01d cuWe&wilY ugldk3WJ 219 v8qu5Be&8J 23.s 19.2 Sduth QOR sh&s 179 vuq0d3q# 25.7 45.9 Onax@alosl&iiiJ ? 13:0 16.1 16.5

10< m

c= 22.3

Cn -1

)
mm

TABLE 7: STATISTICAL ANAUSYS PERFORMED fiiiODIFIED (AAE=A <=10 AM Auumr St8adh8 9.1 ME
SD Author 10< AFI c= 22.3 SD 9.8 M-StdiOS 10.2 S.1

CORRELATIONS

ST-*

~--:=~

<=1OAFI

....

Ba6.oilvi8cdtY Ondm8tdos ViMsy wdanmmdosM-hbcdi M-KarWm@io Ausmr M-E#mglhJuk

ISD

17.4

12.6

lAutbur lREdWabkk

] M-hhtmdjo

I 10.0

4.0 I 3.4

I
o

I M-~

6.0 7.2

q API

m(w)

I Pr(glh ) lw9dmB)l

Pe(jldD)

(q) 1 Vd (c@ I Vo(Cp) I ?%6

I
1 2 3 4 s 6 7 8 9 10 11 12
13 14 15 16 17 18 19 m 21 22 23 24 2s 26 27 28 w 30 31 32 33 34 35 36* 3P 3~* 39 40 41 42 43 44 45* w 47 43 49 so 51* 32 530 ~.

7,1 75 82 82 9,0 10,2 10,s 1Q6 1Q6 11,2 113 113 lN 126 1*O 13.0 13,0 120 1%0 13,6 14.8 IZ8 14# 14# 14,8

M,8
148 1s.0 1%0 1$1 ISJ M,O 160
164 173 179 l&o 16,2 192 19.0 19,0

19,0 19.0 19.6 no 21,1 213 21,3 213 21,7 21,7 21,8 21,8 22,0 22,0

2120 moo 218,7 218.7 117.3 211,1 ls4# 2120 212,0 1s4,8 21Q2 1s4,8 2048 20&o 212,0 212,0 2120 2120 2120 215.4 2113 211,3 211,3 2113 2113 2113 211,3 2113 1%0 207,s 207.3 211,3 211.3 2120 193.1 18Q0 10QO lnll 118.0 149.0 217.4 163.4 163,4 193,8 180,0 190.0 179,6 179,6 179,6 m 17Q1 201.0 177.8 181,0 193.8

4911.0s 4s6s,86 4376.24 2671.64 471s.23 28SQCU 4766,01 476&ol 28s0,04 4s89.07 28sQln 4S09,W 480s.18 49q73 -73 4993.73 49W.73 4993,73 428138 4281s 42a138 42a138 428158 4281s 428138 4281>8 4281.S8 4976,21 4281,S8 4281.S8 4281,S8 4281S8 3723.18 497831 6400$2 4978,21 1182,08 6ss7.26 1807,20 1807.20 3373$6 5333,8s 497821

I
82,17 2426 144.02 786.70 208.7s 189,87 34s>s 91.66 72S,W 37692 202.09 3q76 m,63 m in.42 m 47$,63 493.4s 768@ 2q7s 228,41 2!q31 315@ 419,00 444,10 46s,64 753J7 46%75 19s,s9 IQ1O 19138 16s,82 21sgl 143@ 73,62 26Q11 24733 xq14 lwg3 IIm 15207 229,46 2s&44 W&m 197,81 437.10 2s2s6 7w2n) 83206 18921 378@ 18s.88 S21,38 412.s1 4944s1

W&3
121,83 79627 499.%73 2364.1s 193433 2s7446

2783 4s1,s m =4 334Q0 12S.8 93,0 148.8 1002 111,7 11120 1M6 103,1 393 6Q6 37.4 75.0 753 S3,4 w 413 4Q0 Q7 373 39.8 403 683 563 5S.6 37,0 3%6 w 703

1247 36Ql 773 2Q1 41,7 160 769 93 1s.0 420 17.2 37.7 73 302 17,4 119 17.0 7,2 19.6 16.2 1S6 10,3 9,1 9.6 8.1 %3 142 47 21.8 18.9 153 13.8 1%7 4Q0 3,4 203 %6 80.6 233 to 13,7 11.7 7,6 5.2 2,0 7.8 23 22 41 1,7 S,l 13 1.7 1,9

2283

1Z7 21.0 S28.o

m.n 4993.73 28s8.74


2319.19 2A73,83 2261.17 4993,73 11s1.62 28s674 4993,73 28s8.74 4993.73 16SQ36 Z61,17 28s&74 28s&74 28S8,74 4281,38 4281.s8 428138 4281.s8 711.13 1763.69 1834,76 177839 1991,40 1038,49 663,14 2A17$W 24x@2 6829 797.72 668.63 72s.20 1393#3 180720
101238 main 2062,47

mo 543 159 527 32.7 w

2104,S3 %0,16 374733 91830 313634 142,41 2S602SI

12.9 161 12,8 12J 12,0

%89,48 3430,S0 497821 4978.21 4978.21 I

75.0 I*4 13Z0 IQ 1s,0 40 I&o So@ 7.4 S40 ao 119 229 72 23.8 243 161 122 1S,6 9,6 8,1 S2 14,2 7.1 3098 235 %6 18,3 19.0 m 44 2$4 3.6 N6 220 161 15,6 11,7 9,6 7,7 23 7,8 3.6 22 5.4 2.1 66 1.s 23 23

t
(ME-A
o~

TABLE 9: St&tiOd RcdbofVii10Vcd@h wmge AbR&MG%SD=Stmd Dw., M= Wditkd)

Vod

w
19,8

bllOr
<=10

lwabO@lbKlrw8Ll
2i,417,1

Vo A&LAdi
59

SD
10-22.3 AM%

13 w M-EgbO@ MxmtOdm WKmlOdm 5,7 35,4 20J3

al

8
s s

! R

w q

Si

m mocna)

so

Fw3:SoMionOOILbatcurdaWI
u

hldadure

/
q q

/
m u

q q 9

*
a. o

/
s .,. -----. -

Fiisbdkmldaqrawdy

hlllitmtm q lmaaiddw08

659

=s

IoD.-

m.,

Fg

8Dd.dvi8cdlyspraatwcfk

Ma

E
MB

kc

Fig.9~dkdy,

batcud8tKu

hnliilaahm

m
.
F@ 12 Wkdumtd

F% 11: undauWdoilVi&dy,

bdcamekthfiatulilmilm

ii

cilviumdly,pradmlrk

10-22.3

API R8mge <10

-s-ding
Stsndins 10 ATers~e 1s absolute error 20 25 30 o 5

t
Pro-ant
10-22.3

W ork Vssqucz-Beggc

API Ramge

?rssomt W ork

<10

10

Is

20
q bs.lmte
q

25
rror

30

35

Average

Prosont W ork

IO-22.3
Vasquez-Be~ga

API R~nge
roscmt Work

c 10
Vasquoz-B*, gs

10

15

20 Arersce

25 qbsolate

30
qrror

35

40

4s

50

F~17b0tbrmde0mwdb@

~~~~-

661

Proscnt
10 223

W ork Bgbogah J-ok

API R80C0 Proseat


<10 =-L--.I. -5 --=-l... b ----

W ork

10

20 Average

30 qbsoluto
qrror

40

50

Fw18D60&trlm6c0aly~~

dlld

rOOub

10223

API
Komga <10

10

15

20
qrror

25

30

Avoraco

sbs.lute

Ftg 19 GHaumtdcdlWcOmtycOnddm.~

0fbt6t -

Prcsoat 10223

W ork
K

qrtoatm

odjo

API Romgo < Iu

rcaoat W ork

1
0 2 4 A~orsgo 6 sbsolato 8 10 12 14

qrror

You might also like