Atnmpheric

Enuironment Vol. 11, pp. 837-845.

Pergamon

Press 1977. Printed in Great Britain.

OPTIMUM DESIGN OF VENTURI SCRUBBERS

K. C. GOEL Advance Engineering Branch, Atomic Energy of Canada Limited, Chalk River Nuclear Laboratories, Chalk River, Ontario, Canada* and K. G. T. HO~LANDS Department of Mechanical Engineering, University of Waterloo, Waterloo, Ontario, Canada (First received 24 May and in final form 27 December 1976) A~~ct-Althou~ venturi scrubbers are capable of removing submicron particles, they do this at the expense of a high pressure drop in the gas stream, with a coincident high fan operating cost. Designs of venturi scrubbers which, for a given gas flow rate and particulate removal e@ciency, have minimum pressure drop, may be called optimal. This paper considers design procedures for optimized design. On the basis of a straight duct model, the optimum combination of liquid flow rate and gas velocity at point of liquid injection is treated. Using this combination, and using a multi-duct analysis, the optimum distance upstream of throat where nozzles for injecting the scrubbing liquid should be located is treated. Optimized design charts, based on air-water systems and pneumatic ato~~tion are presented. An important result of this treatment is the prediction of very low pressure drops, if an atomization system giving the optimum droplet size (rather than the larger droplet size produced by conventional pneumatic atomi~tion) is used.

1. INTR~DU~ION

submicron particles, it does it at the expense of a very large pressure loss in the gas stream, with a consequent high fan cost, particularly in running costs. Con~quently it is highly desirable to minimize the pressure loss across venturi scrubbers. The development of op~~zed venturi scrubber designs which have ~nimum pressure loss has awaited the development of a proper quanti~tive theory of their operation. Recently such theories have begun to appear: Calvert (1970); Boll (1973); Hollands and Goel (1975); God and Hollands (1974), to name a few. Although these theories are based upon a number of assumptions which are open to question, they mode1 the performance of full-size indust~al scrubbers reasonably well, provided proper care is taken in the venturi design. In particular, care must be taken so as to ensure that the scrubbing liquid is uniformly ~stributed across the venturi throat; and the internal aerod~amics of the venturi must be such as to avoid separation of the flow from the walls. These theories now appear to be adequate for obtaining at least nor-op~mum designs, if not fully optimal. The literature contains little info~ation on the optimum design of venturi scrubbers, although the existence of optimum combinations of liquid flow rate and throat velocity which minimize the. pressure loss was expe~imen~lly observed by early investigators (Eckman and Johnston, 1951; Byrd and Dewey, 1957). The complexity of determining an optimum venturi geometry was pointed out by Uchida and Wen (1974).
Al~ough the venturi is capable of removing

The object of this paper is to present an optimization scheme based upon modem theories. Because of the extreme complexity of the problem (due largely to the multipli~ty of design variables) it was found necessary to make some simp~f~ng assumptions. The most important of these is that for the purpose of determining the optimum combination of liquid flow rate and gas velocity, the venturi is modelled as a single straight duct. On the other hand, for the purpose of deter~ning the optimum dimensions of the real venturi, a multiduct analysis is used From these considerations a procedure for designing a near-optimum venturi scrubber is described An interesting result of this study is the high potential for a much reduced pressure loss if an atom~tion procedure with a controllable droplet size is used, rather than the pneumatic atomization presently use& 2. STRAIGHT DUCT MODEL OF A
V~TURI SCRUBBER

* Present address:
Canada.

CNS,

Ontario

Hydro,

Toronto,

The physical model used to obtain the optimum conditions which should prevail at the point of liquid injection, namely the optimum combination of gas veIocity, liquid flow rate and drop diameter, is shown in Fig. 1. The dust laden gas enters a straight duct of (~eoretically) infinite length and having a radius equal to the radius at that point of liquid injection of the real venturi. Liquid is assumed to be introduced in the form of droplets of diameter, d. In the first analysis of this optimization problem and droplet diameter d will be assumed independent of the gas velocity and liquid flow rate. Later, in the paper, consideration will be given to the case where pneumatic atomization is used, in which case d is fixed by these variables. The droplet velocity at the point of liquid

837

K. C. GOI:L and K. Cr. T. HOLLANDS

---

DROPLET,

DIAMETER

Fig. I. Straight duct model of a venturi scrubber injection will be assumed to be zero. The other assumptions regarding the flow and fluid properties are the same as used commonly by previous investigators and include the assumptions of one-dimensional, ihcompressible and adiabatic flow, that of uniform liquid distribution across the cross-section, and small liquid volume fraction. Inertial impaction is assumed to be the dominant particle collection mechanism. To further simplify the problem. we will neglect for the moment the pressure loss due to wall friction. However, this simplification turns out to be fortuitously acceptable in this instance since it compensates to some extent for the recovery due to kinetic energies of the gas and the liquid which is achieved in the divergent duct of the real venturi. Due to this reason and the fact that most of the momentum exchange and particulate collection of a real venturi take place in the throat region (where a zone of high relative velocity between the phases exists), Calvert (1970) could achieve a remarkable success in comparing the results predicted by such a simplified model with experimental data on real venturis.
3. GOVERNING
EQCI;ATIONS

The target efficiency expression used is that given by Goel and Hollands (1974) :

where qn is the target efficiency at the point of liquid injection. Equation (4) represents a correlation of Walton and Woolcocks (1960) data as given by Calvert (1970) and its use allows for the flexibility of using as accurate a value for qa as possible. Let us for the purposes of the present study use Calverts (1970) expression for q., i.e.
K

VU = i K, + 0.7

1.

The differential equation for particulate collection is the same as that used by Calvert (1970). Integrating this (assuming the liquid velocity at the point of injection to be zero) for a straight infinite duct (see Goel and Hollands, 1974 for details) and making use
of Equations (2-4, we get the following expression for the total number of collection units in a duct :

The pressure loss for the physical model described above is given simply by the change in momentum of the liquid from zero to its value when the liquid velocity is equal to the gas velocity. Thus: (A& = M IV&. (1) where

2,44

I
,..

2-t 3Y
:+ .~. -

,ytan-

N,.T = ~-

24 0. ..c ?? + ---- (1 + 0.15 Re'.') B

R%

vy

-1

(6)

In order to obtain an expression for the number of collection units, we use the drag coefficient expression given by Hollands and Goel (1975) as: C, = c,, (u,/PJ~5. (2)
and may be given by using Equation (5) as:

(7a)

Equation (2) approximately represents the standard curve over the Reynolds number range of practical interest here. The drag coefficient at the point of liquid injection, CD,, can be prescribed as accurately as desired. Let us describe it by a close fit to the standard curve as given by Dickinson and Marshall (1968): CD, = 0.22 + $1 r? + 0.15 Rf$&CJ.(> ). (3)

y =

!g = ._h:W.
N

C,.Pp&qa

W-3

It is now useful to define a dimensionless group X by the relation : x = C,p,r;&,d;/~i;. X is independent of the drop diameter. Substituting Equation (8) into Equation @I (6). there

Optimum design of venturi scrubbers

results :

839

= 9.1 M

* 00)

Equations (1) and (10) give respectively the pressure loss and total number of collection units in a venturi scrubber due to the simplified model used. These equations will now be used to obtain optimum values of drop diameter, liquid flow rate and gas velocity.

4. OPTIMUM DROP DIAMETER

It is interesting to note from Equation (1) that the pressure loss is independent of drop diameter. Therefore, the optimum drop diameter is simply the one which maximizes the number of collection units. Figure 2 exhibits the dependence of (~~,~jM) on X and Y as given by Equation (10) where qa is related to Yby Equation (7a). It is seen from this figure that for each value of X, there is a unique value of qU which corresponds to maximum (N,,&f). The peaks of the curves in Fig. 2 are plotted in Fig. 3. Since X is independent of drop diameter, the value of ~,_,,,, in Fig. 3 can be used to obtain an optimum value of the drop diameter for any set of the other variables on which q. and X depend. The physical meaning of the existence of an optimum drop diameter for maximum efficiency is explained below. It is known that the target efficiency of a droplet increases by decreasing the diameter and increasing the relative velocity between the gas and the droplet. However, when two droplets of different diameters are introduced simultaneously into a gas stream (with the same initial relative velocity), then the smaller droplet is accelerated faster than the bigger one with

0101
Fig. 2. N&M

04 %

06

0.6

1.0

plotted as a function of X and )I~.

the result that the relative velocity goes to zero earlier in the case of smaller droplet (which is a more efficient particulate collector) than in the case of the bigger droplet (which is a less efficient particulate collector). Consequently, there exists an optimum drop diameter which is given by Fig. 3. The feasibility of using Fig. 3 for practical purposes will be discussed in the next section.
I OPTIMUM COMBINATION OF GAS VELOCITY AND LIQUID FLOW RATE

It is well-known that both the collection efficiency and the pressure loss of a venturi scrubber are increased by increasing either the gas velocity or the liquid flow rate or both. The problem of practical interest is then to find out those combinations of gas

Fig. 3. Optimum value of target efficiency at the point of liquid injection.

AX. 11/9--D

840

K. C. GOEL and K.

G. T. HOLLANDS The problem is to find optimum values of d, c,,~ and A4 which will give minimum pressure loss. Fat this purpose, first a value of usa is assumed, then the value of X is obtained from Equation (9) the value read from Fig. 3, the value of d,,, calculated of %Wpr by using Equation (7). the value of Z read from Fig. 4, and the corresponding value of (AL/?)~ calculated from the definition of Z. The value of M corresponding to the assumed value of lsD is obtained by using either of Equations (I} and (10). The procedure is repeated for severat assumed values of t:ra to find out a feasible value of U$~. The results of this calculation are summarized in Table 1. The last column in Table 1 gives the drop diameter produced by pneumatic atomization calculated by using the Nukiyama- Tanasawa (1938) equation. for the values of I,, and M shown in the table. Table 1 shows that the optimum drop diameter and liquid flow rate increase with the gas velocity, whereas the drop diameter given by the Nukiyama-Tanasawa equation (sometimes referred to in this paper as the N-T equation) decreases with gas velocity. Thus although it appears that the pressure loss can be decreased (at Ieast hy~theticaiIy~ to a very low value by decreasing the gas velocity, there are certain Jimitations in doing so. The most important limitation is quite obvious and is due to the difficulty in obtaining proper atomization, for it is seen from Table I that the optimum drop diameter required at low velocities is considerably smaller than that which would be produced by pneumatic atomization, as predicted by the Nukiyama--Tanasawa equation. This then strongly suggests the desirability of devising a separate atomization method which will produce a controlled droplet size, smaller than that produced by pneumatic atomization. One possibility is to use pressure atomization such as that used in, say, ejector vcnturi scrubbers (Harris. 1965). (In this case the energy requirement of the pressure atomizer will also have to be taken into account in the optimization pro-

velocity and liquid flow rate which for any desired eNiciency will minimize the pressure loss. For this purpose, eliminate M from Equations (1) and (lo), and non-dimensionalise using Equation (8) to get:

It is seen from Equation (11) that the dimensionless parameter 2 = (A~)~~~/(~~,~~~~~C,) is in general a function of two parameters, namely, X and k: However. if it is assumed that the droplet diameter is independent of gas velocity. then for any value of X, the optimum value of Y may be given from Fig. 3 and Equation (7a). Thus combining Fig. 3 and Equations (7a) and (1 I), we get Fig. 4 where the dimensionless parameter 2 is plotted as a function of X. In other words. Fig. 4 gives the value of Z (which is directly related to pressure loss) for any value of Q, f&e. X) used, provided the drop diameter is given its optimum value as predicted by Fig. 3. It appears from Fig. 4 that to get minimum pressure loss (i.e. minimum Z) one should use as small a value of gas velocity as possible. However, in practice, a value of X less than a certain value may not be found feasible. To illustrate this, the following numerical example is chosen. Let the following data be specified: I .3 kg/m3 (properties of air) 1.X x WPa.s 1000kg/m3 (properties of water) 0.072 N/m 2800 kg/m3 1.0 x IO-rn 4.6 (99% efficiency).

Fig. 4. 2 plotted as a function of X using optimum drop diameter as given by Figure ?

Optimum design of venturi scrubbers Table 1. Illustration for optimum design r@. m/s 15 30 45 60 90 X x IO3 2.528 10.1I 22.75 40.44 91.0 0.165 0.839 0.872 0.895 0.917
dw Z Wh M dN-r eqn.

841

/Im 53 68 79 84 98

x 10s 2.21 4.36 7.80 10.14 20.0

in HZ0 3.62 6.96 12.4 16.2 31.9 3.08 I .48 1.17 0.86 0.75

pm
326 163 109 82 55

cedure.) The other limitations of the above method of reducing pressure loss by using lower gas velocities, are (i) the large amount of liquid required (as seen from Table I) and (ii) the requirement of a longer venturi (a fact which is completely overlooked here due to the simplified model used for optimization) at low gas velocities.
Dependent drop diameter

it was assumed that it is unaffected by the gas velocity

Having discussed a hypothetical means of minimizing the pressure loss in a venturi scrubber, let us now examine Table I from a more conventional point of view. i.e. assuming pneumatic atomization. The optimum drop diameter and the drop diameter predicted by the Nukiyama-Tanasawa equation as given in Table I are plotted as functions of arrain Fig. 5. This figure shows that the curves intersect at a well defined point (corresponding to agO = 57.5 m/s in this case). This point may be mistaken for the optimum operating point for a scrubber for which the NukiyamrTanasawa equation is valid and which is designed to operate according to the specifications mentioned previously. However, a deeper thinking will reveal that this is not the case since in obtaining optimum value of the drop diameter for maximum efficiency,
400

usC In reality, however, the drop diameter produced by pneumatic atomization must be a function of the gas velocity at the point of atomization or inversely the gas velocity to be used in obtaining X and Y in Equation (10) is in reality a function of the drop diameter rather than a constant as was treated before. Therefore, the derivative of ue,, with respect to the drop diameter must also be involved in the optimization procedure. The actual optimization procedure in this case becomes too involved to be treated analytically. Therefore a numerical search technique was used The algorithm for this is not given here for lack of space but reported elsewhere (Goel, 1975). The results of the computations are plotted in Figs. 6-8 for a mixture of air and water, the combination used most in practical scrubbers. The properties of the air and water are as given in the example discussed above. (Due to the dimensional nature of the Nukiyama-Tanasawa equation, it was not possible to I I , , , , , , , ,

135,

r-

------r-

I
t got

,N-T

equaflon

(see

table

I)

For

maximum

efflclency

I
30

oi__--,_+
15 pa

+7&---i&m/s

4 N C

6 T

IO

Fig. 5. Comparison of drop diameter for maximum efficiency with that produced by pneumatic atomization as predicted by Nukiyama-Tanasawa equation.

Fig. 6. Optimum gas velocity at the point of liquid injection for a mixture of air and water (using N-T equation for drop diameter).

I(. C. COEL and K. G. T.

HOLLANUS

i
0

I 4

I 6

I 8

I 10

1
N C. T

N C

Fig. 7. Optimum liquid flow rate for a mixture of air and water (using N-T equation for drop diameter). present these results in a dimensionless form.) In performing the actual computations, a more accurate drag law, namely equation (8), was applied loc&y for the full length of the venturi, and the integration for Nf,r was performed normally rather than using the cfosed-form expression given by Equation (10). Figures 6-8 depict that in general the smaller or lighter the particulates to be collected or the higher the number of collection units desired, the higher the liquid flow rate and gas velocity must be; the corresponding pressure loss will also be higher. These figures can be used in the actual design of real airwater venturi scrubbers to estimate the optimum

Fig. 8. Minimum pressure loss obtained by using optimum gas velocity (given by Figure 6) and liquid flow rate (given by Figure 7) as predicted by Equation (I 1. values of the gas velocity at the point of liquid injection and the liquid flow rate to be used for any given requirement. However, one must bear in mind that the use of the Nukiyama-Tana~wa equation is implicit in these figures. Figures such as 6 8 can be readily prepared for other gases and scrubbing liquid. and for other droplet size equations.
6.

OPTIMUM VENTURI DIMENSIONS

The optimum values of the gas velocity and the liquid flow rate as given by Figs. 6 and 7 were obtained from a rather simplified model of a real ven-

-)

---.

..~._~

1,

___.-.._

..----.._

__._._c

Fig. 9. Schematic diagram of a real venturi.

Optimum design of venturi scrubbers

843

::_ 0

,
I

,
2 q (or

,
3 1,/r+)

,
4

,/_$
5

os I Fig. 10. Effect of $ on optimum ueJus.

2 1
3 4

-05

L,

,m

turi. It is now postulated that these optimum values of the liquid flow rate and the gas velocity at the point of liquid injection are also applicable to a real venturi. The real venturi is shown in Fig. 9. In general any independent set of six variables will completely define the venturi geometry. One such set which is convenient in the present treatment is given by /?i, P2, lT (total length), I, (throat length), us. and usl (the gas velocities ugs. and ua, may be used in place of radii r, and r, since these can be related directly as the gas flow rate is assumed to be specified.) For brevity, only conical ducts are examined. The angle of convergence /Ii and the angle of divergence /I2 are usually taken for conical ducts to be 12.5 and 3.5 respectively which seem to avoid unnecessary pressure loss due to flow separation. The total length (IT) is assumed to be specified as before. In addition, let us assume for the moment that: 1, = $r, (12)

Fig. 11. Effect of I, on optimum ueJu4. a mixture of air and water. In general, the ratio (u,Ju,,) is a function of five variables, namely Q,, I,, pP di, N,,r and $. The effect of I(/ on the ratio (u,Ju,J is shown in Fig. 10 where the ratio of pressure loss obtained from the resulting venturi to that predicted by the simplified model (given in Fig. 8) is also shown as a function of $ for fixed values of the other parameters indicated in the figure. Since (Ap), given in Fig. 8 is independent of $, the ratio of the two pressure losses plotted

where $, the ratio of throat length to its radius, is fixed. (A common engineering practice is to use the value of $ between 1 and 3.) The remaining variables are ugoand uer.However, as was assumed in the beginning of this section, the optimum value of urrO obtained from the simplified model and given in Fig. 6 is also applicable to the real venturi as well. Therefore, the only independent variable which now needs to be determined is uat which can be easily obtained by trial and error so that the required number of collection units is obtained. This was done by using a comprehensive method for predicting the performance of a venturi scrubber as developed by Boll (1973). NukiyamaTanasawa equation was used for the drop diameter. Typical results are plotted in Figs. 10-14 for

loil~-a5
2 3

2x PPdP

IO

-9

kg/m

Fig. 12. Effect of pP ds on optimum IL&,,,

K. C. GOEL and K. 6.

T. HOLLANDS

ent lengths of the real venturi scrubber. It is thus seen that the pressure loss in a venturi scrubber can be decreased by increasing Its total length. Howcvcr. the length of the scrubber will be limited to different values for different applications depending upon the space available for installation. The maximum length available should clearly he used. Figs. 17, 13 and 14 show that r4,.~,j0incrcasrs with 11~ di and N,., and decreases with increasing Q,. The relatively slight variation in the pressure ratio shown in these figures indicates that optimized designs based on straight duct theory should not bc very far from being fully optimal.
7. RECOMMENDED DESIGh PROCEDl:RE

in Fig. 10 gives an indication of the variation of pressure loss in a real venturi scrubber with tj. It is thus seen from Fig. 10 that the overall pressure loss in the venturi scrubber is relatively insensitive to the value of I(/.Its value can therefore be determined from other engineering considerations such as to avoid an abrupt divergence of flow leaving the convergent duct. Another interesting result is the effects of 1, on the two ratios. shown in Fig. I I. As in the case of $ discussed above, the pressure loss given in Fig. 8 is independent of l7 as well. Therefore, the plot indicates the variation in pressure loss obtained by using differ-

For design of air-water scrubbers of speciticd ,V,..,. Q,. and pr, dj. it is recommended that I,,~and !&I be determined from Figs. 6 and 7. Values of 1);. /lz and $ be decided by intctnal fluid Row considerations, so as to avoid separation of Row from the walls. The total length of the venturi should he decided so as to be as long as possible hut so ;S to meet any external constraints such as height ot room available. etc. (If guidance is needed for thk consideration a graph such as Fig. I I should he prcpared so as to show the pressure drop penalty for short overall length.) The ratio of t;Jc,, necessary to give the required N,,,, should then be detetmmed b) trial-and-error. using a comprehensive multi-duct analysis such as given by Boll (1973) or Gael and Hollands (1974. 1975).

8. CO&CLCSIONS
I. Use of the design procedure given in the prcvious section should yield design of venturi scrubbers with pneumatic atomization which are very close to optimal; i.e. have minimum pressure loss for a given gas Row rate and particulate removal etficiency. 2. If an atomization method can be used in the venturi which produces droplet of :I controllable size (smaller than that produced by pneumatic atomization), chosen to be optimal for the given venturi application, very large reductions in venturi pressure loss should result. Acknowlr~~rmn~ts-This work was supported and fellowship from the NatIonal Research Canada.

I _/-0. -

___---,---

iI

./f

by a yrdnt Cot&ii of

LIST OF NOlATIO\ c = particulate concentration 111 the gas ctrcam. kg/m3 C, = Cunningham correctlon factor to Stokes Law for particulate slip between gas molecules (required only when d, < 1 /cm, dimensionless CD = drag coetficient ol droplet, dimensionless C,,, = drag cocflicient at the point of liquid injectron d = drop diameter, m d, = particulate diameter. m

Fig. 14. Effect of Q, on

optimum uJc,_

Optimum design of venturi scrubbers K = = I= L = m= .w = S,,.,. = (Ap)r = (Ap).r = Q= r = r,, =

Re = Re, =

X45

= c= %a = r10= 11, = = L,,= .Y = Y= Z =

inertial impaction parameter. C, pp dilu, - 0,)!9 p,d, dimensionless length, m liquid flow rate, gallon/(lOOO ft of gas) mass tlow rate, kg/s mass flow rate ratio, = m,/m,, dimensionless number of collection units, dimensionless pressure loss, Pa pressure loss, inch H,O volumetric flow rate. m3!s radius. m radius of venturi at the point of liquid injection, m drop Reynolds number, = lug - u,ld pO/pU, dimensionless Reynolds number at the point of liquid injection, Ir*:,.- r/z,d P~;P# velocity, m/s gas velocity at the point of liquid injection, m/s liquid velocity at the point of injection. m/s rclativc velocity between gas and droplet. rs - I,. m/s rclativc velocity at the point of liquid injection, = rpu - c,,, m/s C, p# p, di t&p:. dimensionless 1:;~. - I. dimensionless (Apb p:&Y,,r pp di C,), dimensionless

p = of the particulate t = of (or at) the throat of the venturi T = total

REFERENCES

Boll R. H. (1973) Particle collection and pressure drop in venturi scrubbers. I & EC Fundamentals 12, 40.-50. Byrd J. F. and Dewey E. L. (1957) Venturi scrubber in odor control. Chem. Eny. Prcg. 53. 447451. Culvert S. (1970) Venturi and other atomizing scrubbers efficiency and pressure drop. 4.f.Ch.E. Journal 16, 392-396. Dickinson D. R. and Marshal1 W. R. (196X) The rates of evaporation of sprays. A.1.Ch.E. Journal 14. 541-552. Ekman F. 0. and Johnston H. F. (1951) Collection of aerosols in a venturi scrubber. Ind. Ena. ._ Chem. 43. 135X-1370. Goel K. C. (1975) Analysis and optimum design of venturi scrubbers, Ph.D. Thesis. Dept. of Mech. Ena.. 1). of Waterloo, Waterloo, Ontario~ Canada. Goel K. C. and Hollands K. G. T. (1974) A general method for predicting particulate collection efficiency of venturi scrubbers. ASME paper No. 74-WAIAPC4. Harris L. S. (1965) Energy and efficiency characteristics of the ejector venturi scrubber. Journal APCA 15.
302-305.

Greek Lerrers /I =

17= q., = $ = /r = p = (i =

angle of divergence, degree target efficiency of droplet, dimensionless target efficiency at the point of liquid injection constant in Equation (12) dynamic viscosity, Pas density, kg/m3 surface tension between gas and liquid, N/m

Hollands K. G. T. and Goel K. C. (1975) A general method for predicting pressure loss in venturi scrubbers. ! &
EC Fundamentals 14, 1622.

Subscripts TV = at the point of liquid injection (I = of the gas I = of the liquid

Nukiyama S. and Tanasawa Y. (1938) Trans. Sot. Mech. Eny. (Tokyo) 4, translated by E. Hope, Defence Research Board, Department of National Dcfence (Canada), Ottawa. Uchida S. and Wen C. Y. (1973) Gas absorptton by alkaline solutions in a venturi scrubber. I & EC Process Des. Der;elop. 12, 43743. Walton W. H. and Woolcock A. (1960) The suppression of airborne dust by water spray. Aerodynamic Capture of Particles, ed. E. G. Richardson, Pergamon Press, Oxford. 129-153.