Analytlca Chrmrca Acta, 229 (1990) 267-276 261

Elsewer Science Pubhshers B V., Amsterdam - Pnnted m The Netherlands

Choice of paraffin inhibitors for crude oils

by principal component analysis
M CRISTANTE and J.-L SELVES *

Luboratorre Instrumentatron et Mesure, Dkpartement Mesures Physrques, IUT, 115 route de Narbonne, 31062 Toulouse Cedex (France)

G GRASSY

Groupe de Recherche en Modklrsatron, 31 allies Jules Guesdes, 31400 Toulouse (France)

J. ORRIT and E GARLAND

DRD P, SNEAP Etoblrssement de Boussens, 31360 Sarnt Martory (France)

(Received 27th December 1988)

SUMMARY

Accumulation of paraffin wax m the eqmpment for petroleum productlon causes large fmanclal losses. The most effectwe and
most economx means ofavoldmg deposItIon IS often to add chemlcaf addltlves to the crude oil, but It 1s not easy to find the most
smtabfe addttlve for a partlcufar sltuatlon because the phenomena mvofved are too compfex for trachtlonaf approaches. Prmclpai
component analysts IS used for data reduction of parameters to represent chenncal addltlves and crude 011s On the basis of
physlcochermcaf propertles (mcfudmg spectral parameters) and the actwty of addltwes on the crystafhsatlon temperature of the
paraffin wax, crude ok are cfawfied mto two t’armhes and chermcaf mfubltors mto three families A modef for predxtmg the activity
of addltlves IS discussed.

The liquid state of petroleum IS probably the lowing causes: (1) a sufficiently low temperature
most important factor in the growth of its use, to induce deposition of wax [2,5-71; (2) accumula-
particularly in the chenucal industry and in the tion of paraffin wax on surfaces m contact with
energy sector. The choice of petroleum rather than crude 011 [&lo] when these surfaces have a lower
solid or gaseous fossil fuels is often decided by temperature than that of the crude 011; (3) surface
this property because petroleum is easier to trans- roughness of pipe walls; and (4) the crude 011
port and store. However, for some types of crude composition [11,12]. With regard to composition,
oil, and m certain condltlons, the liquid state 1s the paraffin weight fraction IS certainly the most
not stable and the transformation of a fraction of important factor. The presence of light com-
it to the solid state brings about accumulation of pounds such as methane, ethane, propane and
paraffin, causing considerable financial loss (stop- butane affect the solubihty of paraffin [7,8] and
page of well production, impossibility of unload- some chemical families like asphaltene, aromatics
ing a cargo, etc.). This phenomenon has been and resins can reduce paraffin deposltlon [8,13,14].
studied for several decades [l-3]; apparently the Another feature is the flow characterlstlcs of crude
oldest bibliographic reference dates from 1865 [4]. oils [10,15]. There is also broad agreement on the
Previous works agree in general about the fol- means of avoiding or suppressmg deposltlon,

0003-2670/90/$03.50 0 1990 Elsewer Science Pubhshers B.V

268 M CRISTANTE ET AL

which can be done thermally (hot oil, steam and was first diluted (1 + 9) with tetrahydrofuran
heaters [16-181) mechanically (scrapers, etc. (THF) and 20 ~1 of this solution was injected.
[16,19]), or chemically with solvents and additives Four columns of CL-Styragel (porosity 105, 104, lo3
(wetting agents, dispersants and wax modifiers) and 500 A) connected in that order, were used to
[2,20,21]. obtain the SEC recording. The solvent was THF
The most economic and often the most effec- at a flow rate of 1.5 mini.
tive of these processes consists of adding chermcal Retention indexes were determined by a rapid
substances to the crude oil. The structure of these pyrolysis thin-layer chromatographic (TLC)
additives is mostly polymeric and their essential method with a flame ionisation detector (FID).
role is to slow down or to modify the crystalhsa- The overall time required for ten samples is about
tion of paraffin [22-241. They improve the flow 45 min. An Iatroscan model TH-10 was used; the
properties of oils and limit the surface deposition FID was operated with a hydrogen flow rate of
of crystals. This is a preventive method; the others 160 ml mm’ and an air flow rate of 2 1 mm’.
are generally more expensive and more difficult Samples were prepared by dissolving about 0.5 g
and are used as a last resort. However, it 1s not of the chemical additive and 0.5 g of n-octadecane
easy to find the additive which gives the best in 100 ml of xylene; 0.6 ~1 of this solution was
results because many additives are available, and applied on a silica-gel II Chromarod (for
their effectiveness depends on the crude oil bemg TLC/FID; Iatron Laboratories). The Chro-
treated [ 181. Moreover, several deposition marods were developed in a TLC tank with filter
[8,9,15,21] and mhibition [6,21,22] mechanisms paper as wick, and xylene/THF mixture (80 : 20,
have been proposed. v/v) as the developing solvent. The rods were
Many variables act upon these systems and the removed from the tank when the solvent front had
phenomenon is too complex to be treated by a travelled 8 cm, and allowed to dry for 1 mm in an
traditional approach. It is therefore, appropriate oven at 60°C, before being placed m the Iatro-
to use multivariate data analysts and particularly scan model TH-10. The variable R f was calculated
principal component analysis (PCA). The results from R, = average distance travelled by additive/
of PCA will help in data reduction, in the classifi- distance travelled by n-octadecane.
cation of crude 011s and chermcal inhibitors, and The crystallisation temperature (variable ACD)
m prediction and modelling. of the chemical additive was measured by using
differential scanning calorimetry. The apparatus
was a Mettler TA-3000 and the cooling rate was
EXPERIMENTAL 10°C min-’ between 9O“C and -30°C. Be-
tween these temperatures, this parameter could
Chemical addltwes not be measured for the three compounds 11, 17
Twenty-four additives were studied. The first and 22 and was given the value zero.
treatment was to extract the active compound The mass percentages of carbon and hydrogen
from the commercial product by evaporation. were determined with a Carlo Erba 1106 elemental
Various methods were then applied to determine analyser, and were used to calculate the value of
the necessary data. the variable C/NO as the ratio WC/(100 - %C
The solubility of the additive in n-octane at - %H).
25 o C (g 1-i) was measured (variable SOL). Two parameters were calculated from the ‘H-
Size-exclusion chromatography (SEC) was used NMR spectra (Brucker WP80 spectrometer):
to obtain data related to average molecular weight RMNl = Intensity at 2.1 ( k 0.1) ppm/intensity at
and structure (variable MW). These results and a 1.0 (kO.5) ppm
set of polystyrene samples of known molecular RMN2 = Intensity at 3.8 ( + 0.3) ppm/intensity at
weight were used to calculate MW. A Philips 1.0 (kO.5) ppm
Pye-Unicam PU-4015 apparatus with a differen- Six parameters were calculated from 13C-NMR
tial refractometric detector was used. Each sample spectra. These are defined by the ratios between
CHOICE OF PARAFFIN INHIBITORS BY PRINCIPAL COMPONENT ANALYSIS 269

the peak heights for carbon in particular func- such compounds have dispersant properties for
tional groups (located by DEPT sequences): paraffin deposits.
C3 = -CH,-O/CH,; C, = CH,,,,,, x lO/CH,; The crystalhsation temperature (variable
TCCB) was determined by differential scanning
C, = CH x lO/CH, ; C, = O=C-OR/CH, calorimetry (see above); the cooling rate was 10 o C
C7 = O=C-OR/CH,; C8 = O-C-OR/CH; mm’ between 90 o C and - 40 o C.
Four parameters were obtained from infrared
C9 = CHJCH,; Cl0 = CH/CH,
spectra (see above) by calculating the ratios be-
Seven parameters were extracted from infrared tween the heights of absorption bands located at
spectra (Brucker IFS-88 spectrometer). They were particular frequencies:
obtained by calculating the ratios between the
heights of absorption bands located at the follow- II = A,,,/&,,; 12 = 4,,/4,,,;
ing frequencies (cm-‘): 13 = &,,/A,,,; 14 = 4,,/4,,
IRl = &,,/A,,, ; IR3 = A,,,,/&,; The width of bands were: 1460 f 5, 1600 f 10,
965 + 10,770 k 5, and 720 + 10 cm-‘. Because no
IR5 = Ann/A,m ; IR6 = 4,x/&w,;
band appeared nearby, 1975 cm-’ was chosen for
II/s = 4,m/&n; IA/5 = 4,,,/4,,,; the horizontal baseline.
1315 = &s/An,2
The widths of the bands were 1730 k lo,1461 f 5, Measurement of activrty
1253k13, 1175+14, 1372*6, 1121 +ll and The crystallisation temperature (T,) of the
2850 it 2 cm-‘. Because no chemical additives ex- paraffin content in a crude oil mixed with 2000
hibited a band nearby, 2000 cm-’ was chosen for mg kg-’ additive was determined by differential
the horizontal baseline value. scanning calorimetry [12,25-271. This temperature
corresponds to the formation of the first paraffin
Crude oh crystals, which point can be determined, as their
In order to investigate the degrees of effective- appearance is accompanied by an exothermal ef-
ness of the additives, it was necessary to test their fect. Measurements were taken with a differential
activity on different types of waxy crude oil. With microcalorimeter (Mettler TA3000) with a cooling
the aim of classifying each crude oil, the following rate of 10 o C mm’ between 85 o C and 0 o C. In
analytical characteristics were determined. order to preserve a homogeneous vanable, the
The weight percentages of paraffin (vanable activity (additive ability to lower the Tc) is ex-
Par) and of heavy paraffins (variable C30 + ) that pressed by the ratio between the Tc of the pure
possess a carbon atom number greater than 29, crude oil and the Tc observed with inhibitor. The
experimental results are given in Table 1.
were measured by gas chromatography. The heavy
paraffins are liable to produce deposits that are
difficult to eliminate by scraping. The internal
reference method (with the normal paraffin PRINCIPAL COMPONENT ANALYSIS
C,,H,,) was used with a Carlo Erba HRGC-5160
gas chromatograph. The capillary column of Principal component analysis (PCA) mvolves
Chrompack SIL-5B was 25 m long; the carrier gas the description of similarities and differences in a
was hydrogen; FID was used. The temperature set of objects (additives) described by N variables
programming was 50°C for 3 min, 3.5O C min-’ (physicochemical properties or activities on the
to 15O”C, 5°C min-’ to 300°C and 300°C for T,). This set of objects is in N-dimensional space;
40 mm. projections of the objects on particular factorial
The weight percentage of asphaltene (variable planes which contain the major part of the mfor-
As) was determmed by a standard method (Norm mation about the variability of the system are
NF T60-115). Some authors [8,13,14] consider that done by PCA [26,28,29]. Graphical representa-
270 M CRISTANTE ET AL

TABLE 1
Actwty on the crystalhsatlon temperature of mne crude 011s

Addltlve Tl T2 T3 T4 T5 T6 T7 T8 T9
1 1 10 0.97 1 10 125 1 10 1 19 1.10 1.02 100
2 108 0 96 104 1 20 1 10 1 12 128 0 93 1 55
3 0.92 1 03 104 107 1 10 154 1 39 1.45 2 58
4 0.97 1 10 1.08 1.27 122 2 06 1.52 2 21 2 21
5 1.00 100 1.08 1.33 113 1 85 1 68 1 55 1 82
6 1.08 0 97 1.08 1.20 1.03 106 1 19 1 00 1.19
7 108 0 88 1.10 1.17 1.01 1.09 123 0 93 1 35
8 0 93 0.99 108 1 27 1.11 1.54 1.60 1.68 2 58
9 0 92 0.97 100 1.00 0.96 0.93 1 .oo 1.00 1 19
10 0 97 0.89 100 1.05 1 01 0.90 1.00 1.00 1 41
11 0 93 0.89 1.00 100 0 98 0 95 1.00 0.95 1.07
12 1.04 0 89 1.08 1.30 1 01 123 1 23 1 35 129
13 0.95 105 1.06 1.30 1 16 1 85 168 2 10 3 10
14 0.98 0 88 1.02 1.00 0.96 0.88 0 94 0 98 1.55
15 0.97 108 1.08 1 36 0.96 168 2 00 2.10 2 58
16 100 0.96 106 1.27 1.18 2 31 1.60 2.10 2 58
17 0 97 0.98 100 102 0 98 0 93 1 03 0 93 103
18 106 0.87 1 12 1.33 103 1 16 1.19 108 100
19 100 1 10 1 16 1.39 113 1 85 1 78 2 10 2 07
20 0.95 0 89 1.08 1.22 0 98 0.90 1 14 102 1.07
21 1.02 0.88 108 1.15 101 109 107 100 1.19
22 0.93 0 88 100 1.05 0.98 0.95 0.97 0.98 107
23 100 0.91 1 06 1 13 1.05 1.03 1.07 1.05 0 91
24 0 88 0.95 1 10 1 33 1.13 1.32 2.13 2.10 3 10

tions on the principal factorial planes allow the Selectcon of the physlcochemlcal parameters of
similarities between the objects to be seen and the addltwes
provide information about their differences; if an In order to select explanatory variables for the
object is very close to others in these planes, their PCA method, the correlation matnx, which con-
information contents are similar according to the nects physicochemical parameters to the activittes
descriptors used [30,31]. Essentially, PCA, is used on the Tc of different types of crude oil, was
to classify a set of objects. The program used calculated. Then these parameters were inserted
comes from a software package called MOLDE-
SIGN, kept at the ANADO library, R.I.,
C.N.U.S.C. in Montpellier (France). These pro- TABLE 2
grams have been used for other applications PCA of crude 011s
[32,33].
Elgenvalues Percentage Total
percentage
Classrfkatron of crude orls
1 4.306 53.82 53 82
Table 2 gives the results of the PCA method
2 1402 17.52 71.34
applied to a set of 59 types of crude oil defined by 3 0901 11.26 82.60
8 parameters. It can be seen that the first three 4 0 568 7 10 89.71
eigenvalues contain more than 80% of the total 5 0.450 5 62 95.32
mertia of the multidimensional data; the first two 6 0.182 2 28 97.61
7 0148 1 85 99.46
eigenvalues hold 71% of the information, and their
8 0043 0 54 10000
eigenvectors define the PC plane.
CHOICE OF PARAFFIN INHIBITORS BY PRINCIPAL COMPONENT ANALYSIS 271

IR5

IR6
IR3
* 19 RMN3

c4
RUN 1 c5 SOL
. 18
TZ

77 1315
. 16 c3
___,R,____________._______........_.. .:. . . . . _ . . . . ____.______..............
T5 Cl0

c7
Rf
ACD

C6

. 14 ‘NO

T3
14/5 .

Fig. 1. Correlation circle.

. (.Iaxis 2
____.....___.__....__._.._._...____________________....________..........
. 16 924

n 3 . 13
. 19

84
Ii73 IRS
. 15 11 x
!

r 17 x

22 x

/IR6
.____,........_._......______................. 9 x _..................... __..)

axis 1

14x x10

21 0
60

20
18 0
lo 70
20 0

23 o
I...............................

Fig. 2. PCA of the physicochemical properties of the additives: principal plane.

272 M CRISTANTE ET AL

BXlS 2
. . . . . . . . . . . . . .._....._____..._.....__.._.________.___________
./. . . . . . . . . . ..____...________.___________.______
n 3 9x

11 x

22 x
. 24
. 13 17 x
10 x
14 x
ma

n 4 . 15

. 16

. . . . . .._......____._._.._______.____________._~5 ‘siih ..

0 20
o 23

n 5 144
13
I
rl 02
0 21

. 19

0 12
06

07

01
0 18
. . . . . ..______.__....._._...____..........__._...________._____
. . . . . . . . . . . . . . . ..___...__________._______.
Fig. 3. PCA of actwtles on. Tc prmcipal plane.

into the correlation circle [34-361 so as to simplify number of variables without seriously modifying
their comparison (Fig. 1). the representation of the system. There is some
Investigation of these results shows several fea- redundancy between parameters, e.g., RMN3 with
tures. The parameters C3, C4, 0, C8, ClO, ZR6 (correlation coefficient r = 0.92), C7 with ZRl
C/NO, RMNl and SOL seem to have little effect (r = 0.90), and C9 with Z1/5 (r = 0.81). The first
on the system. Their elimination can reduce the two redundancies can be explained by the fact
that these parameters pertain to the same struct-

TABLE 3
TABLE 4
PCA of the physlcochemtcal properties of the addmves
PCA of the actwltles on the crystalhsatlon temperature
Elgenvalues Percentage Total
percentage Elgenvalues Percentage Total
percentage
1 4 905 44.59 44.59
2 3.706 33 69 78.29 1 5 579 61.99 61 99
3 0 798 7 26 85.54 2 1770 19 61 8145
4 0.533 4.84 90 38 3 0.668 I 42 89 07
5 0 498 4 52 94 91 4 0412 4.58 93 65
6 0 305 2 II 97 61 5 0.224 2.49 96.15
7 0 122 1 11 98.78 6 0.175 1 95 98.09
8 0 096 0 87 99.65 7 0098 109 99.18
9 0.026 0 23 99 89 8 0.043 0 48 99 66
10 0008 0.07 99 96 9 0031 0.34 100 00
 CHOICE OF PARAFFIN INHIBITORS BY PRINCIPAL COMPONENT ANALYSIS 273

ural properties. The thud correlation suggests that
compounds which have few or no carbonyl func-
tions have relatively short polyethylene chains.
After these removals, 11 explanatory variables were
retained: MW, R,, ACD, IR1, IR3, IR5, IR6,
I1/5, I3/5, I4/5, C6.
Classlfrcatlon of the addrtwes
The results of the PCA of the physicochemical
properties of the additives (Table 3) are repre-
sented by the eigenvalues and the eigenvector
coordinates m the variable space. It can be seen
that the prmcipal plane (Fig. 2) defined by the
first two eigenvectors represents more than 78% of
the total inertia of the data.
Actwity on the crystalhsatton temperature
The results of this PCA can be used to differen-
tiate the additives according to their action on the
T, of the different crude oils. Table 4 gives the set
of data relative to the new variables of the system.
The first two eigenvectors, which hold 81% of the
information, define a plane on which the projec-
tion of every ObJect is represented by a number
and every variable by a vector (Fig. 3). The vector
length is directly related to the quality of represen-
tation of these variables in the principal plane of
this analysis; the longer the vector length, the
better described the variable. The relative bearings
of two vectors can define the degree of connection
between the activities that they represent.
RESULTS AND DISCUSSION
Classrfrcatlon of the crude oils
Figure 4 shows the relation between the ObJects
(crude oils), the variables (analytical features) and
the first two principal axes to which the following
meanings can be given. Axis 1, which represents
54% of the initial variability of the data, separates
asphaltenic crude oils from waxy crude oils. Axis

_.____
_______
____
_____
_.___.__________
_______
i.~~~:.t.. ________
____.
0 0 +
+I +
a, I +
0 +

...

Risk zone

!..........__...__
.......__..................(..I . . . . . . . . ..___.__.______....
Fg 4 PCA of crude oils pnnclpal plane Set text for explanation.
274 M CRISTANTE ET AL

2 essentially distinguishes the crude oil according in then chemical structures (variables ZR3, ZR1
to the variables 13 and 14. These parameters, and Z3/5). Family, B contains the additives 1, 2,
which come from the ratio of infrared IR spectral 6, 7, 12, 18, 20, 21 and 23; these inhibitors have a
bands, probably represent different proportions of relatively large MW parameter and some acrylate
the hydrocarbon families constitutmg crude oil groups (variables Z4/5 and Z1/5). Family C
(aromatics, saturated compounds, etc.). The com- (compounds 9, 10, 11, 14, 17 and 22) comprises
plexity of crude oils is such that it is impossible to the ObJects which are marked by the variables
relate, without ambiguity, the Infrared spectral ZR6, ACD and R,. In this family, which is less
bands to the chemical farmhes which compose the homogeneous than the others, there are some phe-
011s. nolethoxylated compounds (9,17 and 22) and some
Investigation into these planes (Fig. 4) can be ammated additives (10, 11 and 14).
used to define two fan-&es of crude oil according The 24 additives tested represent the essential
to their chemical composition. Family 1 (objects part of the currently available commercial inhibi-
denoted by X) groups together the crude 011s tors; comparison of the basic chemical structures
which have a paraffinic feature (those which will of these three families shows some important dif-
give easier deposition of paraffin are located in ferences.
the risk zone). Family 2 (ObJects denoted by 0)
includes crude oils with a large asphaltene per-
centage. Because of their positions, none of the Actwlty on the crystalhsatron temperature
objects denoted by + can be attributed to either The following meanings can be attributed to
of these two families. the principal axes of this analysis (Fig. 3). Axis 1
With the aim of simulatmg the behaviour of the is a “general” activity axis because the additives
additives towards the set of crude oils susceptible are classified according to their effictency. From a
to creating paraffin deposits, nine crude oils were vector projection on this axis, rt can be observed
selected; they are represented by numbers 1-9 on that the worst represented variables (Tl, T,, T3)
Fig. 1. Four of these crude oils appear m the risk belong to crude oils which are located in the risk
zone, with oils 1 and 2 m extreme positions; the zone in Fig. 4, whereas the variables T6, T7, T8,
other crude oils have a more or less marked paraf- which are the best represented, describe crude oils
finic feature. relatively far from this zone. Therefore, a relation
exists between the position of crude oils m the
Classlftcation of addltwes principal planes of the PCA and the influence of
Investigation mto the principal plane defined their constitution on the activity of the additives.
by the first two prmcipal components makes it Axis 2 achieves some separation into types among
possible to distinguish some farmlies of additives the specific activities of miscellaneous additives.
in terms of their behaviour and their (A structure/activity analysis to find a compound
physicochemical properties. Axis 1 of this plane with the largest range of activity will lead to
(Fig. 2) separates the additives both in relation to special variable issuing from the first prmcipal
their crystallisation temperatures (parameter component.)
ACD) and according to the presence m their The three families previously defined can now
chemical structure of ester groups (parameter C6) be distmguished m the following way. Family A
or other groups (parameter ZR6). Axis 2 essen- corresponds to the compounds which are effective
tially perrmts the differentiation of additives which on almost all the crude oils studied; among them,
possess ester groups by the use of the following the most effective additives are numbers 19 and 4.
parameters: ZR3 (characterizmg acetate groups) Family B contains the additives which have a
and 14/S (related to the acrylate group). specific activity on the crude oils of type 1. In
Three farnihes are clearly distinguished. Farmly general, they seem to be less effective than those
A comprises the additives 3, 4, 5, 8, 13, 15, 16, 19 of famdy A. Family C consists of products which
and 24; all these compounds have acetate groups have little or no activity.
CHOICE OF PARAFFIN INHIBITORS BY PRINCIPAL COMPONENT ANALYSIS 275

A “general” efficiency rate can be attributed to 24), which are now Included in the present model.
each additive on the basis of the study of their Work is m progress to determine some other
coordinates on the prmcipal axis. Thus four in- parameters to describe additives and crude oil.
hibitors of the A family have an efficiency rate of
90% and the rate of the B family lies between 20 The authors thank SNEA(P) for its contnbu-
and 35%. tion to the development of this work by the means
From these results, there is clearly a strong of scholarships and technical assistance.
relationship between the physicochenucal para-
meters representative of the different families and
REFERENCES
their type of activity. Moreover, the diversity of
crude oils used shows that the nature of the crude 1 D Shock, J D. Sudbury and J J Crocket., J. Pet Technol ,
oil can influence the efficiency of some inlnbitors. 7 (1955) 23
2 R S Fulford, Sot Pet Eng of AIME, Pap, 5413 (1975)
Conclwon 171
3 C.A. Bdderback and LA McDougall, J Pet Technol , 21
Principal component analysis was used to study
(1969) 1151.
some crude oils, a set of paraffin-deposition in- 4 J H A Bone, Petroleum and Petroleum Wells, 2nd edn ,
hibitors, and the activity of the additives on the Phdadelphla, PA, 1865, p 26
crystalhsation temperature of crude oils. 5 R.J Cole and F.W Jessen, 011 Gas J , 58 (1960) 87
Crude oils can be classified mto two farmhes, 6 R.C. Price, J Inst Pet., London, 57 (1971) 106
7 J Svetgoff, 011 Gas J., 82 (1984) 79
which are essentially distinguished by their
8 G E. Addison, Sot Pet. Eng of AIME, Pap, 13391 (1984)
asphaltene and paraffin contents. The additives 203
are divided into three families, which are clearly 9 P.A Bern, V.R Wlthers and R J R. Calms, European Off-
distinguished by their physicochermcal properties, shore Petroleum Conference, EUR206, London, October
their structure and their activity. The activity of 1980, Sot. Pet. Eng of AIME, 1980, p 571
10 G P Engelen, Offshore Technology Conference, OTC 3518,
the additives is modified by the nature of the
Houston, TX, Apnl/May, 1979, Sot Pet Eng of AIME,
crude oil. However, objects scattered on the Tc 1979, p 1385
activity plane show that one of these families has 11 R Gdanslu, 31st Ann Southwestern Petroleum Short
the largest range of activity. These different types course, Texas Tech Umverwty, Lubbock, TX, 1984, p 449
of evidence make it possible to clarify the situa- 12 M E. Newberry, J. Pet Technol , 36 (1984) 779
13 S M Bucaram, J. Pet. Technol , 19 (1967) 150.
tion a little and to evolve a process for improving
14 J. Bnant, Rev Inst Fr. Pet, 18 (1963) 1
the choice of an inhibitor. 15 G.A Holders and R. Mills, 11th Austrahan Conference on
With the crystallisation temperature as an ex- Chermcal Engmeenng, Inst Eng Aust , Bnsbane, 1983, p
ample (other variables are being investigated), the 111.
INIPAR model which is proposed can be used for 16 G.T Woo, S.J Garbls and T.C. Gray, Sot. Pet Eng of
AIME, Pap., 13126 (1984)
the multidimensional location of a new crude oil
17 E M Sunkm and A.I. Sergeer, Neft Khoz , 6 (1967) 39
and for the choice of additives. The determination 18 J-L. Mendell and F W. Jessen, Sot. Pet. Eng of AIME,
of the physicochemical parameters of a crude oil Pap., 2868 (1970) 171
permits its location on the crude-oil principal plane 19 K.M Fagm, Pet. Eng., 1 (1945) 105.
m PCA. When the crude oil family is known, use 20 M.E Newberry and K.M. Barker, Sot. Pet Eng of AIME,
Pap., 13796 (1985) 53.
of Figs. 3 and 4 pernnts a rational choice of the
21 J M. McCall and R.L. Johnson, 31st Ann Southwestern
additives to be used. Petroleum Short Course, Texas Tech Umverslty, Lubbock,
After every new experiment, the crude oil can TX, 1984, p 457
be introduced into the analysis in order to make 22 J C Petmelh, Rev Inst Fr Pet., 34 (1979) 771
the INIPAR model evolve steadily, by increasing 23 T.R Slfferman, J Pet Technol , 31 (1979) 1042
24 J C Petmelh, Rev Inst Fr Pet, 34 (1979) 791
the number and types of the crude oils. The same
25 P Claudy, J.M. Letofe, B Chague and J Omt, Fuel, 67
process can be carried out with new additives. (1988) 58.
Initially, the set contained 22 additives and this 26 F Cadlez and J P Pages, Introduction i I’Analyse des
model was applied with two new additives (23 and Don&es, Sot. Math. Appl. Scl Hum., Pans, 1976, p 221
276 M CRISTANTE ET AL

27 F. Bosselet, J.M Letoffe, P. Claudy, B Damm and P. 32 G Grassy and M. Bonnafous, Chemom Intell. Lab. Syst.,
Maldonado, Thermochtm Acta, JO (1983) 49 5 (1989) 221
28 S Wold, K Esbensen and P. Geladl, Chemom Intel1 Lab 33 G. Grassy, J C Teulade, J P Chapat, M Slmeon de Buoch-
Syst, 2 (1987) 37 berg and M Attlso, Eur J Med. Chem -Chum Ther , 17
29 0.M Kvalhelm, D W Aksnes, T Brekke, M 0. Elde and E (1982) 109
Sletten, Anal. Chem , 57 (1985) 2858 34 G. Mdhaud, J.N Talbot and G Coutns, Theraple, 29
30 P. Broto, G. Moreau and C Vandycke, Eur. J. Med Chem (1974) 611.
-Chum Ther., 19 (1984) 79 35 0.M Kvalhelm, Chemom. Intell. Lab. Syst., 2 (1987) 127.
31 J de Lagarde, Imtlatlon ?I I’Analyse des Donnees, Dunod, 36 E Dlday, J. Lemalre, J Pouget and F Testu, Tr2utement
Pans, 1983, p. 85 des DonnCes statlstlques, Dunod, Pans, 1982, p 190.