Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
.,;
.J C-PTbS-03-o{
Advances in
Estimating Gas wen Deliverability
..!
By A. G. WINESTOCK and G. P. COLPITTS"
,.
(16th A.nnllal Technical Alee-ting, The Pel1"oleum Society of C./.ll!., Calgary, 11fay? 1965)
flow test, where the flows at each rate are consecutive, from the Houpeul"t equation can be checked by means
the deliverabilit}· estimates often are not reliable_ of simulated field data generated with the gas-well
The new drawdown technique is not limited to use in
gas wells. It can also be used to estimate well produc- simulator, whose reservoir properties are knmirn_ The
ti"it:y and diagnose wellbore damage in oil or water ensuing two sections of the paper then use this pro-
systems. cedure to explore the accuracy of the conventional
method of drawdown test analysis and present an im-
proved drawdown analysis technique, referred to as
the normalized method. Follo,ving this, a recommended
field testing procedure is outlined, in terms of which
the current testing practices, such as isochronal and
flow-after-flow tests, are evaluated_ Finally, appli-
INTRODUCTION cation of the normalized drawdown method to oil and
,"vater wells is discussed brief!}r.
AS flow in a porous medium is differen~
from
G oil or water flow in that it does not obey the
Dai'cy Law. In the gas-reservoIr-engineering litera- THE DIGITAL GAS-WELL SmULAToR
-, ture (I), this difference is now commonly called "tur-
bl:llence_" Turbulence occurs because the low viscosity The gas-well simulator is a digital computer pro- ,,. ..
of gas (compared to that of oil or water) results in gl-am which simulates the flow of gas through the
correspondingly higher gas velocities under a given reservoir and into a well. The program solves the non-
pressure gradient. At these higher velocities, the gas linear difference equation for one-dimensional un-
suffers inertial energ,Y losses as well as the usual steady-state gas flow in a porous medium. It also
viscous euergy (Darcy) loss_ This energy loss changes solves the pressure-rate relationships which describe
th,~ relationship between gas veIocit}r and pressure flow in the production string. Its main features are
gradient from linear, as in Darcy flow, to non-linear. listed below:
Turbulence is greatest in the vicinity of the wellbore, (a) _ Flow in the reservoir is single-phase, unsteady-
where velocities are large_ Thus, it gives rise to an state, radial and horizontal.
extra pl-eSSUl'e drop around the well - an effect simi- (b). Permeability and pOl'osity can vary radially
lar to that of wellbore damage. throughout the system.
The dependence of gas viscosity and density on (c). Flow includes turbulence, and the turbulence fac-
pressure introduces a second non-linearity into gas- tor can vary radially throughout the system, .'
flow theor}r_ At anyone time, these parameters will (d), Gas viscosity and density are continuous func-
vary throughout a reservoir across which pressure tions of pressure.
gradients have been impressed. In addition, the pres-
sure at any point will change with progressive de-
pletion. "Impe>';al Oil Ltd., Calgary, Alta.
(e). 'Veil radius and external radius of drainage can CONVENTIONAL DRAWDOWN METHOD
be specified.
(f) . Tubing and casing dimensions and production- The conventional drawdown method yields a per-
string lengths can be specified. meability-thickness product. It is based on the fact
( g). A large variety of well-boundary conditions can that, by Equation (1), which assumes a constant rate,
be imposed in succession, as follows: a plot of P,:: - p",/ versus log t results in a straight
line. From the slope per logarithmic cycle of this plot,
(i) Constant or variable bottom-hole rate the permeability-thickness product can be calculated:
(ii I Constant or variable bottom-hole pressure
(iii) Constant or variable wellhead rate, including zero
mte (wellhead shut-in with after-flow) md·ft (21
(i \' ) Constant or variablE! wellhead pressure
(\') Flow through a wellhead critical-flow prover of spe-
cified size. where In = slope per lo~aritillnic c~'c1c from fl plot of (n~ p..,2 \
vc;, In,C;loL
(h 1. BOllndal"}' conditions at the external radius of
drainage can also be imposed in succession, as The kh product is an important reservoir para-
follows: meter. Its role in determining deliverability and the
degree of wellbore damage or improvement is dis-
(i) Constant or variable reservoir I"ate cussed later.
(ii) Constant or variable. reservoir pressure.
Equation (1) has been written for a well producing
Simulator cases were nm on both the IBM 7090 at a constant rate. In practice, it is often diffic.ult to
at the University of TOl'onto and the IBI\'l 7044 at the establish a constant rate during a flow test. As shown
Calgary Data Centre. later with the conventional drawLlown method, a vari-
The gas-well simulator accurately approximates one- ation in rate of only a fraction of a per cent through-
dimensional radial gas f1m\', but it provides no simpli- out a test can result in highl)r erroneous kh values,
fied algebraic expression of pressure-flow-rate rela- because the variation being considered in Pl~ - p.j
tionships. One such simplified expression is described can be of the same order of magnitude. The next sec-
in the following sectioll. tion describes a new drawdown analysis method which
a voids this difficult~·.
In 1959, Houpeurt (2) presented an equation that To account for variations in rate, we developed il
pl"Ovides an empirical basis for calculating reservoir new analysis which we call the normalized drawdown
characteristics from well tests. The Houpeurt equation method.O\< It is as simple as the conventional method,
incorporates turbulence in an approximate solution has been ~uccessfully field-tested, has be~n verified
to the unsteady-state radial flow of gas into a well on the ga~-well .simulator and (:~ln be justified theo~
producing at a constant rate. As originally presented, retically.
the equation describe.:; gas density at the ,'vellbore.
The normalized drawdO\vn method i~ ba::;ed on Equa-
Expressed in terms of pressure J the relation is
tion ({]) J which describes the drawdown in a gas
' ..' ell flowing at a varying rate. (See App{mdi:r for
theoretical discllssion).
This equation assumes that the system is homo- + Bq:!.: (tl + additional terms I';')
geneous, horizontal and of uniform thickness. It also
assume~ that several pl'essure-dependent properties of
[n Equation (3), the flow rate, q~(t), is the in-
the gas (i.e., viscosity. compl'essibility, average pres- stantaneous flow rate rather than some average COIl-
sUL'e) can be l"epresented by some suitably averaged stant value for the test. The additional terms are usu-
constant values. ally negligible, provided that the changes in rate are
not excessively rapid_ For example. when flowing II
The time-independent nature of the turbulence well through a critical-flow prover, the additionl\l
term, Bq",~, in Equation (1) deserves special comment. terms are generally .small. \Vhen these term~ are
The term may be included in this form when the rate neglected, Eqlw-tion (3) is eqUivalent to treating qo:
throughout the region of turbulence has reached the in the Houpeurt equation as the instantaneous flow
\Vellbore rate. As Houpeul't (3) points out, this gen- rate rather than the average rate. Treating the rate
erally happ~ns quickly, because most of the turbulence as a variable makes it nece~sary to "normalize" Equa-
occurs within a very small region around the well, so tion (3) to obtain a meaningful straight-line plot ver-
that a pressure disturbance can usually mo",,'e through sus the logarithm of time; i.e., Equation (1,.):
it in very little time.
It ~hould be pointed out that the turbulence term al- p,2 _ p",! ttl - Bq2~ (U
1 --
II
·1 TABLE I
,'.
"
COMPARA:IVE ERRORS BY THE TWO DR.~WDOWN lvlETEODS
~
i
Per "enl Err;r in kll !J)'
-_._-------------
B f(Fi~ -;)2"rh - (-;~ --:;2u.r)d (q, -q,) N ;TIn'Jlized Jlfethod
!J.!,:zT psia:! - - - - - - - - - - - - - - Actual ------ Bqlt2
Run -l,h
- (~j2 -02"r12 q, kl, Cotweu 1iJf1ul With Wi(ll!JlIl
Number (M,d/D)' J:er celt :e~ cent mj-ft J\lel!l:;d T!trb'/lence Turbulence p;.2_ p 2I<"f
----
, 1.. .... .0381 0.723 0.51 0.18 400 60 - 2 11 0.85
J 2 ...... .0194 0.246 0.30 0.20 400 81 -10 20 0.89
5."
, 3 ......
4 ......
5 ......
.108
.0763
.285
0.0683
0.0321
0.0225
3.3
5.8
3.9
2.1
0.72.
2.0
120
120
40
88
19
67
6
6
8
16
7
10
0.23
0.10
0.045
6 ..... .404 0.00766 5.~ 1.8 40 3-1 9 9 0.008 .' .
7 .....
8 .....
.00808
.0158
0.00213
0.00100
5.1
5.5
0.07
0.24
1.200
1,200
4
9
3
6
3
6
0.051
0.023
"
9 ...... .0381 0.0226 4.0 0.36 400 18 9 9 0.099
10 ...... .0194 0.00768 4.9 044 400 15 5 6 0.19
11 ..... .108 0.00213 -1.4 1.2 120 38 8 7 0.018
12 ...... .0763 0.00100 7.3 0.36 120 12 9 8 0.008
13 .. .285 1.13 0.72 2.3 40 '165 5 34 0.36
14...... .404 0.3B3 4.6 3.2 40 90 6 16 0.20
15 ...... 00808 0.106 0.85 0.04 1,200 1 - 6 - 2 0.81
16 ...... .0158 0.0500 2.0 0.[4 1.200 -2 -9 - 5 0.73
,." .'
~
tional and the normalized method.
1;:1 1637 tLJ;z'T' ."t To assess the relative accuracy of the two meth-
1, 1= m' m .. -l (5)
cds, sixteen drawdowns were run on the gas-well sim- "
/
precede the instantaneous rate is incorporated into ,
CD
the additional terms. At any given moment, the effect N'i ""
of the difference between the current rate and all pre-
~
, 0-
N ._
ceding rates will have been attenuated! provided that ~
~ ::,------- j---tJ-l--c---
1 +-1-I---------ll-----j method. Depending on the choice of slope, the cal-
eulaled kh product is either 39 or ll8 rod-ft. If
~ '0 1 J-~_I-"-LLIf---j-----t---t-j
this were interpreted in the usual mannet·. the lat-
er time Rlope, corresponding to 118 md-ft, would be
chosen as indicating the undamaged productivity. As
:.'i" '. 1.1 the well's actual productivity is not as high as the
'I Ii
,
i .
1
, ., 10
I1H md-ft. would suggest. ami because the annly~ig
inclicate.s damage, the conv~ntional method says that
this well is a prime candidate for stimulation.
TURBULENT 6 ( F")
TOTAL ll,IP") Figure 4- is a plot of the same data, using the
normalized method. In this case. there is only one
Figure. 2.-}:..'r/"(Jr.o; in Calculating k1l from Dra.wdowfls. slope, corresponding to a kh product of 11 md-ft. If
the normalized method without turbulence is used, a~
In Figure 2, the errors of the conventional method shown in Figure 5, the resulting kh product is 13 md-
are cumpared to the errors of the normalized method. it. Thus, the ,veil's low productivity really result~
As explained earlier, the turbulence group, B, may from a low formation kh and not from damage.
not be knuwll. F'igll1'e 2 also shows the errol'S of the
normalized method, but with turbulence neglected. An independent check on these results is pl'ovided
The elTor~ are plotted as a function of the degree of by data from a pressure build-up test. The build-up
turbulence in the s~!stem. The degree of turbulence is yields a kh product of about 12 md-ft., which is in
the Quotient of BQI'~ and p,:! - P"I~. line with the values calculated by the normalized
method.
Figure 2 demonstrates that the normalized method
is a distind improvement over the cunventional meth- A SOUND FIELD TESTlNG PnoceDUllE
od_ As can be seen, the c.onventional method can be
very iIHlccunLte throughout the entire range of turbu- Historically, the AOF of a gas well has been an
acc.epteu index of its Quality. However, a~ a meum;
lence. Both theory and simulator runs show that when
for estimating performance, the AOF plot hali limited
the flow raiL declines throughoul a test, and this is
utility and can be misleading. It is limited in thnt
commonly the case, the kh values calculated by the
it does not indicate damage, only the well's produc-
conventional methoL.l tend to be optimistic. The norm-
tivity in its current state. Furthermore, the indica-
alized method is generally much more reliable and, as
tion of current productivity that it does give call
the line shows, there is some correlation between
be misleading in two respects. First, the true line is a
per cent error and degree of turbulence_ 'Vhen the
curve rather than a straight line, so that the extra-
sixteen runs were analyzed bJ.' the normalized method.
polation i.s in error; second. the position of the line
but with turbulence neglected, the enol'S in kh. al-
on the plot changes with the degree of stabilization.
though still significantly smaller than those of the
conventional method, are generally larger than ,,-hen A more general approach to estimating delivera1>i1-
turbulence is considered. ity is to determine formation kh, skin (wellbore dam-
, , I /4
kh =118 mdfl.'- m
•
.. I ~
12 0
~ 0
~
a
m
, 0
N
IO~
a
. 163?PqzT S-
N
a ";:;: I 0
~
kh = m, ~ ~ U m
0.
, , 1 g ~ o.U)
, 8~
l(
'" 0
~ , ~
10~-----'---;;:20;:;------"'5';:0:-------;;;'0:::0:----=--!20g
10 20 50 100 200
TIME (hours) TIME (hours)
TABLE II -.
..
EXAIvIPLE CALCULATIONS FOR ACTUAL FIELD DRAWDOWN TEST t·
~-- .
Pi = 3,962 psia B - 0.0637 psia'/(Msef/D)'
age) and turbulence from ''l.. ell tests, as will be out- For finite-acting systems! expressions analogous to !.
lined shortly_ These parameters can then be used to Equation (1), but suitably modified, can be developed.
calculate deliverability for varying degrees of wellbore To calculate formation kh, skin and turbulence, at
damage or improvement. In addition, deliverability least two drawdown tests should be run, each begin-
can be calculated as a function of time and can thus ning from essentially static initial conditions. The
provide estimates of both short-term and long-range basic procedure for determining these parameters is
performance. These deliverability calculations can be described by Carter, Miller and Riley (6), but has
made with Equation (1) for infinite-acting systems. been expanded here to include the normalized draw-
down method:
I
l'<n ~';onvenlionol) 39 8 1/8 mel fr.
recorded throughout.
5.-The turbulence group, B, is calculated from
Equation (7):
10
0/ 20 50 100 200
B ~ (n.' - p",,'j / Q,l -
ql':l
(Pi' - p~,,'j / c"
qg:!
(7) ;-,
nally~. '"''Ie should point out that in Equahon gas-,,,,'ell simulator. Curve B shows that the simulated
(7) both the numerator and denominator in-
I ' ....eIl's AOF curve reverses. as did the actual well
volve a difference in t\VQ terms. If the terms curve. As a further check, an isochronal test wa~
entering either of these differences are of run on the simulated 'well, with each flow period be-
approximately the same size, the accuracy ginning from static initial conditions. Curve C rep-
of the calculated turbulence group maJo' be lim- l'C:ll:;ents this case and shows no reversal. In less ex-
ited. Rates can generally be chosen to minimize treme cases, the flow-after-flow curve may look reUH-
this problem. onable. exhibiting no rever:->al, but may yield an in-
G.-Using the value of B determined in step 5, correct slope and a misleading AOF.
the well'g kh product is calculated by the To analyze flow-after-flow data for kh product,
normalized method for both drawdown tests damage and turbulence would mean using the prin-
from Equation (5). ciple of superposition, In this case, the results would
7.-Finally, a value of skin is calculated from be less dependable. Furthermore, as some matching
Equatio'n (1). technique ,...· ould probably be required, results would
8.-l\'Iore than two drawdowns can be run, as de- be less directly obtainable than when calculated from
sired. However, each dra'...· down must start an isochronal test.
from essentially static conditions.
9.-Frequentl.v. at least one flow test is run to DRA\...·DOWN TESTING IN OIL AND '~lATE[1. 'VELLS
stabilized conditions (i.e.. long enough to al-
low the pressure disturbance to reach the In this section, we ,·... ill show how the normalized
external radius of drainage). method can also be applied to oil or water wells to
achieve more reliable results than yielded by conven-
The above testing procedure is a form of isochronal tional methods. The deYelopment will parallel the
AOF test, so that it is already embodied in current earlier discussion of gas-well {.!t·awdown tests - i.e.,
testing practices. In fact, the four-point isochronal it will deal first with the equations and method COIl-
test. with shut-ins between each flow, can provide four ventionally used, then with the modification required
drawdowns and three build-upg; results, therefore, when the rate is changing and, finally, with the re-
can be replicated and thus confirmed. However, pres- sulting normalized method, including an example in-
sure and rate data have to be recorded throughout ,"olving minor rate changes.
the tests. rather than just at the end of each flow
pedod, Equation (8) describes the pressure drop as a
function of time at a well producing oil or water at a
THE FLOW-AFTER-FLD\'V TEST cO'ilstant rate from a homogeneous hurizontal sy~tem
of uniform thickness. The fluwing fluid is slightly
One current practice in gas-well testing is to flow
compressible, has a constant viscosit~l and, becau::le
the well at four consecuti"e rates, with no shut-in
tllL'bulence is negligible under virtually all opeL'ating
time between flows. This procedure, usually called conditions, obeys the Darcy Law.
the flow-after-flow test. can give misleading results.
This is illustrated in Figm'e 6. Curve A shows ~'"
p. - [l"i = 1..J-1.2 l~~ [112 (In to + O.809l + Sj (8\
AOF curve from a flow-after-flo,...· test on an actuClI
well which \\'as producing dry gas and no water. "Vith Here. q liquid [Jaw r:Ite, res 13/D
the first change to a larger flow prover, the rate in- to dimenc;ionle~s time = r,t/r,,:!
creased as expected. Hmo;rever, with successive changes 0.00li33k
dj[[usi\'it~r =
to still larger flow proYers, the rates decreased. To ((liLC
,'erify that this reversal effect could result from con- ~ liquid visco<;itr. CD
secutive flows, a flow-after-flow test was run on the c efrcctivc compressihility. HJ1/vol/psi
and all other terms are as defined previously. A plot
of Pi - p.. . r versus 10glUt yield::; a straight line, and,
from its slope, the permeabilit)·-thiclmess pl'oduet CUll
be calculated:
kh = 162.6 ~- md-fL
Ill"
Figure 6.-CompuTison of Flow-Alter-Flow and where m; is the slope per logarithmic cycle. This is
Isoc!l1'onal Tests. the normalized method.
'-Iii':
-- -ljL
~" =I-~
" --.---- :I,'i,!I1 0
III i'I I III
• , ,
~._--~~ .
I~ 1'1,1: ,Ill
,
"I I , -j
, , I' : ,
IiI ' III
" •
I_--f'-i-!-I-+1-,-+'+H!~t!-H
I
I 0
'I I ,i I, ': ilill --5
,- I I :: IIl+Jjj!f---J--+H+++H++H+++tI
,; I I
~I<T '4~1--+I-1--i-I-:'-\-I++i'-l\!11,;+ft\+H--pf-iCALcuLATEO,IrI1=90md fI
.~
"
''---r: -''-~-;-;--' i"1~,~====:;-~:'-:' +:1i'im
lll - III! II 1,611111
;r·
II .'
-;:; 100,
~
I I ! I·
BlI'---t----~--t_'~-':'P---
I I II
1./
I
1,1
II -:
----- - -1-':
I:I
. ,",II I
I
1Wl<>I---'~'~/~2.LU+++++H+!+Jjj
~::.'- .:
,
....
I-+--i- I -+1+++:+1+1+
1
I I I '
ii 1
~!i;f'~
60
I I I ' Ii '-----'-- I' " :'1 1-+-+-,l-++++++++Wr+Hf---,AC Vrf"f'rf,'{ ruA
:
•
,
1
I I : I
1
:'I_--:-1--+1.--.--.;--_..J'--,-1c-'-+'-j1I'i---
I I' I i 1:1'
.'---,.
I
.,-'1h--',,,,II,,I'j
I , , I I I' :!:,
,,'
,,,I---+--i-I+I-+++++lf
" I I
1 I I I III l lL",~L-L-'-'
I
II lLlw".
1I--,I-l1.Lll111.!!.l.! (::
TIME (rninule~J TIME (minulul
Figw"e 7.-Conventional Drawdown Plot JOT a. Figw'e 8.-Nol'11"zalized DTawd01.0n J.llethod for (],
Simulated Oilwell Drawdown. Si1nulated Oilwell Dra1vdown.
ACUNOWLEDGMENT
Figures 7 and 8 are plots of a digitally-simulated
oil-well drawdown test analyzed by the conventional The authol'S wish to express their appreciation to the
and normalized methods, respectively. Although the management of Imperial Oil Limited for permIssion to
. rate varies only slightly, its effect is so severe that present this paper, and also to Mr. D. P. Bossler for his
many valuable suggestions. They are also indebted to
the conventional plot is not amenable to analysis. The Mr. R. V. M. Zahar for his contributions to the theoretical
normalized method, on the other hand, yields a kh development_
product of 90 md-ft, whereas the actual value used
REFERENCES
in the simulator is 100 md-ft,
DISCUSSION OF RESULTS
(1) Ka.t~~
F.,
I!. L.,
VU?UJ J_,
Conwll, D., Kobayash.i R., Poe~!,ma.nn.
Elenbaas, J., and Wemaug, C., Hand-
J
APPENDIX
HE conditions under which the normalized dra'w- Before integrating (A-G) term by term, we prove
T do·".-n method is valid can be determined using
slightly compressible flo,\, theory_ These results can
the following b.y induction:
A = 70.6 fl
where
kh
\Ve assume that the flow rate, , q (.), can be de-
scribed 01' at least closel.r approximated by a poly-
nomial of the fOl'm
n e-~ In
... ~ a,-:i '" (_II''''J-I-I U-ll~
1=0 m! j = t
J
Before evaluating the integral on the right side of c-'
--ch
Equation (1), it is convenient to write q(.) in the x
form
n and the prouf of (A-7) is complete.
q(-:) = ,~ .;r,(t - -:t i
i=o Thus. the integral in equation (A-61 becomes
By expanding equation (A-3) in powers of T and
eqlwting coefficients in expressions (A-2) and tA-3),
we obtain the system of (n+1) equations for the
(n+ll unknowns, a.,:
'Z" = (-Il"a n
= (_1)"-1 (a,,_l + nta ll )
~J ;.'L
]
:7
11 _1
1=0
-
.:I,(t- -:I'e
T
"
1=
-</t
_e__
CZa i=1
n
~
[
;'1'"
L
"
j=l
i
~ (-l)i+j(j-l)! (t/~)j
1 ~
(A-10)
Or. using the log approximation to the Ei function
and employing standard nomenclature,
:'
'Ve can rewrite this as a pO\ver series in
lows:
~/t as fol- Pi -p"r(t)
q(t) ~ 14~,;" [iOn tD + 0.809) ] (A-23)
:;;: .
T1 = e-~/t ~ (_l)iH (~/t)'-I £ (j~i)! lZjti (A-ll) which is the normalized drawdown equation for
lZo i=l j=i J.
.' \Ve want to know under what conditions T. will be
slightly compressible Darcy flow.
It is noteworthy that the preceding general argu-
small with respect to the first term of equation (A- ment is considerably strengthened' for the special
9). case of an exponential-like decay in rate, such as often ~ ..
The coefficients of the power series are Occurs in practice (e_g., producing a gas well through
" .-,/t n a critical-flow prover). Another special case of in-
Ci = (_1);+1 e_ ~ (j-i)! -t' (A-12)
•r
J.
IZJ terest is that of a well producing at a constant pres-
lZD 1=i
sure drawdown in an infinite system (constant ter-
As ~/t ~ 0.01, a sufficient condition for the con- minal pressure case). The flow rate decreases slowly
-i, vergence of the series in (A-ll) is
C, ::; 1
but continualJy and, over a sufficiently long period,
the percentage change in rate is large. However,
,
"
(_4-13)
despite the large total change in rate. the normalized
Vie can write e-~/t.=:::::: 1, agam because ~/t :::; 0.01. drawdown method yields an accurate slope on the
\ T. will then be small if the term drawdown plot. This agrees with the theoretical dis-
~ cussion which indicates that restrictions on the ap-
n plicability of the normalized method depend on the
(j-i)! . ti
:E .r lZJ (A-I'I) functional form of the varying ra te and not simply on
=i J.
the total rate variation.
is small.
For i = I, Tt' is less than the fractional variation
in rate throughout the test, 'which is less than unity.
For i ~ 2, a sufficient condition on T { is
(A-IS)
where
(A-I6) ; ,. "
Condition (A-15) will be met in practice for a large
variety of functional forms for q (t). In such cases, ~ .
the first additional term on the right-hand side of '"
(A-9) will be small with respect to unity, and hence
with respect to -Ei ( - I;/t).
Now we examine the second additional term: WINESTOCK COLPITTS
n ( _l)i+l
T. ~ Ei(-,/t) ~ (A-l7)
Alvin G_ Winestock received his B.A.Se. degree in enginee.-
- a" i! ing physics from the University of British Columbia in 1958.
i = 1 He joined Imperial Oil Limited upon graduation and worked
We can rewrite (A-17) as as a research engineer in their Production Research Labora- ';
: ...••.
;: •..-! •
r- :
!".- -:
Technology, July - September, 1965, Montreal 119