Vous êtes sur la page 1sur 6

Powder Technology 129 (2003) 53 58 www.elsevier.

com/locate/powtec

A note on wall effect on the terminal falling velocity of a sphere in quiescent Newtonian media in cylindrical tubes
R.P. Chhabra *, S. Agarwal1, K. Chaudhary
Department of Chemical Engineering, Indian Institute of Technology, Kanpur, 208016 India Received 29 November 2001; received in revised form 5 March 2002; accepted 16 May 2002

Abstract The settling velocity of a sphere in a quiescent fluid is significantly reduced by the presence of confining boundaries. The severity of this phenomenon (wall effect) is strongly dependent on the geometrical and kinematic parameters. In this work, the simplest case of a sphere settling on the axis of a long cylindrical tube is considered. The functional dependence of the wall factor on the sphere-to-tube diameter ratio (k) and the sphere Reynolds number has been examined. The wall factor is a function of k only both at very low and at very high values of the sphere Reynolds number (Re). On the other hand, it depends on both k and Re in the intermediate zone. The values of the Reynolds number marking the transitions from the viscous to intermediate and from the transitional to fully inertial flow regimes have been identified as functions of k. Using these criteria to delineate experimental data, the relative performance of some of the available expressions for the estimation of wall factor has been evaluated against a large body of experimental results. D 2003 Elsevier Science B.V. All rights reserved.
Keywords: Wall effects; Wall factor; Sphere; Terminal velocity

1. Introduction It is now well established that the confining walls exert an extra retardation effect on a spherical particle settling in a liquid. Conversely, the terminal settling velocity of a sphere is known to be lower in a confined medium than that when no walls are present otherwise under identical conditions. The effect is caused by the upward flow of the liquid, and intuitively, the narrower the gap between the sphere and the wall, the more pronounced is the effect. There are situations in process engineering applications when a knowledge of this effect is required for a rational understanding and/or interpretation or correlation of experimental results. Typical examples include falling ball viscometry, hydrodynamic chromatography, membrane transport and the application of electrical fields to augment the rates of heat and mass transfer and of separation in multiphase systems. While the severity of the wall effect depends upon the size and shape of the confining walls, the case of a sphere falling at the axis
*

of a cylindrical tube is the most commonly used configuration, both from practical and theoretical standpoints. [1 3] Thus, the ensuing discussion relates to the case of a sphere falling axially in a cylindrical tube filled with a quiescent Newtonian incompressible liquid. Other pertinent studies concerning noncircular boundaries and/or off-center sedimentation of spherical and/or nonspherical particles are summarised elsewhere [2,3]. One simple and convenient way of quantifying the wall effects in the present situation is to define a wall factor f as: f V Vl 1

Corresponding author. Tel.: +91-512-597393; fax: +91-512-590007. E-mail address: chhabra@iitk.ac.in (R.P. Chhabra). 1 Present address: Department of Chemical Engineering, University of Florida, Gainesville, FL 32611, USA.

where V is the terminal falling velocity of a sphere (diameter d, density qs) settling along the axis of a tube of diameter, D, in a liquid of density (q) and of viscosity (l), whereas Vl denotes the unbounded velocity of the same sphere in the same liquid in the absence of walls. Dimensional considerations suggest the wall factor, f, to be a function of a Reynolds number, Re, and of the sphere-to-tube diameter ratio, k = d/D, provided that the tube is sufficiently long for the end effects to be negligible. Thus, one can postulate: f /k; Re 2

0032-5910/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 2 - 5 9 1 0 ( 0 2 ) 0 0 1 6 4 - X

54

R.P. Chhabra et al. / Powder Technology 129 (2003) 5358

Most of the previous pertinent studies have endeavoured to establish the functional relationship postulated in Eq. (2) [3]. It is generally agreed that the wall factor is independent of the Reynolds number both at very low and at very high values of the Reynolds number, while in between these two limiting behaviours, the wall factor is a function of both k and Re in the intermediate transition regime [1,4 10]. Thus, it is customary to identify three flow regimes akin to the Moody diagram for friction in pipes, namely, viscous, transition and inertial/fully turbulent regions. A cursory inspection of the pertinent literature and the reviews on this subject [1 3] suggests that though much has been written about the wall factor and on the relative performance of the analytical/ empirical expressions applicable in the viscous and fully turbulent regions, e.g., see Refs. [1 3,10 12], very little is known about the critical values of the Reynolds number marking the limit of the viscous flow regime and the onset of the fully turbulent region [9,12]. Similarly, it is somewhat surprising that though three to four independent sets of experimental results and an extensive numerical study on wall factor in the intermediate Reynolds number regime are now available in the literature, no cross-comparisons have been made to check for the internal consistency of the experimental results and/or the validity of the relevant theoretical results. This brief communication thus has two objectives, namely, to identify the limiting values of the Reynolds number for the delineation of the flow regime and to evaluate the relative performance of various correlations applicable in the intermediate region.

cate data points. Admittedly, many other investigators, e.g. see Ref. [3], have reported results on wall effects for spheres falling in circular and noncircular cross-section vessels, but sufficient details have not been provided for the values of these results to be recalculated in the format consistent with the present definitions. Some investigators [13 15] have reported their results in the form of a deviation factor from the Stokes law. Evidently, their so-called K ( = CDRe/24) factor includes the combined effects of inertia and of confining wall. The values of the corresponding wall factors were evaluated using the procedure outlined by Uhlherr and Chhabra [9] in conjunction with the following expression due to Khan and Richardson [16] for drag on a sphere in the absence of walls:
0:31 :06 3:45 CDl 2:25Re 0:36Re0 l l

This empirical expression was stated to be applicable in the range 0.01 V Rel V 3 105, thereby covering the entire range of conditions covered by the data summarised in Table 1. Finally, a comment about the experimental accuracy of the values of wall factor is also in order. It is readily conceded that though the experimental determination of f can entail maximum errors up to 12 15%, the mean errors are of the order of 5% in the creeping flow region and about 10% at high Reynolds numbers [8,9]. Therefore, the socalled 5% and 10% criteria will be employed to delineate the critical values of the Reynolds number marking the onset of the intermediate and of fully turbulent regimes in this work.

2. Data base 3. Analysis and discussion In order to cover wide ranges of conditions, data has been culled from as many sources as possible from the open literature. A summary of the range of conditions encompassed by various studies together with the other relevant details is provided in Table 1. Clearly, a wide range of conditions is seen to be covered, and altogether, there are about 1260 data points including several repeat and dupliAs mentioned previously, the wall factor when plotted against the sphere Reynolds number on a semi-log graph paper yields a typical S-shaped curve as has been shown elsewhere [7,9,11]. While many expressions of varying forms and complexity are available for the estimation of f in the low Reynolds number region [1 3], the following

Table 1 Range of conditions covered by data Source Range of k Range of Re or flow regime Viscous 0.03 318 2.45 10 4 991 0.012 19,080 8 10 4 0.076 0.02 1369 0.0376 47,110 0.0224 186 turbulent Number of data points Viscous 30 1 1 2 29 2 6 7 0 Fully turbulent 0 20 38 218 0 41 136 12 35 Total 30 49 91 492 43 60 356 105 35 Symbol in figures n 5 o 4 4 x +,

Lee [13] McPherson [14] Engez [15] Fidleris [24] Iwaoka and Ishii [10] Chhabra and Uhlherr [8] Uhlherr and Chhabra [9] Kehlenbeck and DiFelice [23] Bogus and Stamatoudis [25]

< 0.168 < 0.16 < 0.99 < 0.85 < 0.90 < 0.53 < 0.95 < 0.89 < 0.65

R.P. Chhabra et al. / Powder Technology 129 (2003) 5358

55

theoretical expression due to Haberman and Sayre [6] has gained wide acceptance in the literature: f 1 2:105k 2:0865k3 1:7068k5 0:72603k6 1 0:75857k5 4

Though initially, it was thought to be applicable up to k c 0.8, subsequent numerical calculations [17 19] show it to do a good job of predicting the value of f up to k c 0.9. Suffice it to add here that Eq. (4) is in excellent agreement with experimental results also [2,3], and hence, it will be used here for the prediction of f in the so-called viscous regime. On the other hand, at very high Reynolds numbers, the wall factor f again becomes independent of the Reynolds number. Detailed comparisons between the data for circular and square cross-section fall vessels and available empirical expressions clearly show the following two expressions to be most reliable [9,11,12]. Newton [20] put forward the following analytical expression for f in this regime: f 1 k2 1 0:5k2 0:5 5

This study covers the range of conditions as 0.08V k V 0.7 and Rel V 200. Clearly, Eq. (7) is implicit in f, and hence, an iterative procedure is needed to calculate the value of f for known values of k and Rel. Suffice it to add here that Eq. (7) reduces correctly to Eq. (4) as the value of the Reynolds number is progressively reduced. Other than this theoretical study, DiFelice [22] put forward the following correlation for the wall factor f in the intermediate Reynolds number regime.  a 1k f 9a 1 0:33k where a is related to the terminal Reynolds number as: 3: 3 a 0:1Rel a 0:85 9b

Subsequently, Kehlenbeck and DiFelice [23] proposed another correlation as follows: 1 kp f 10 1 k=ko p where both ko and p are functions of Rel as follows: ko 0:283 :524 0:041Re0 l 1: 2 ko For Rel V 35,
:434 p 1:44 0:5466Re0 l

Though no details are available about the derivation or the limits of applicability of Eq. (5), it is assumed to apply over the complete range, namely 0 V k V 1. On the other hand, based on his own experimental results (0.11 V k V 0.83), Munroe [4] presented the following empirical correlation: f 1 k1:5 6

11a

11b

and for Rel>35,


0:8686 p 2:3 37:3Re l

11c

In the common range of applicability, the predictions of Eqs. (5) and (6) differ only marginally and thus, Eq. (5) will be used here to describe the dependence of f on k in the fully turbulent region. In the intermediate Reynolds number regime, some theoretical work is available which has been summarised and reviewed elsewhere recently [3]. However, only Wham et al. [21] have presented sufficient information for their results to be recalculated in the form required here. After some rearrangement, their final expression can be recast in terms of the wall factor, f, as follows: f 1 0:037080:5Rel A C 1 0:037080:5fRel B 7

They also reported that only slightly larger errors are incurred in the prediction of f if a constant value of p = 2.2 is used in Eq. (10) and the corresponding value of ko is obtained from the expression: ko 0:27 :65 0:05Re0 l 1: 2 ko 12

where A 1:5844 0:1016lnRel B 1:5844 0:1016lnfRel C 1 0:75857k5 1 K k 2:0865k3 1:7068k5 0:72603k6 8a 8b 8c

Eqs. (10) and (12) are based on the experimental results of Fidleris and Whitmore [7] and Kehlenbeck and DiFelice [23] and are thus limited to k V 0.85. It is rather surprising that neither the theoretical predictions of Wham et al. [21] have been contrasted with experimental results nor the applicability of Eqs. (9a) and (9b) or Eq. (10) has been demonstrated using independent experimental data other than those used to evaluate the empirical constants, like p, ko, a, etc. This will be attempted here. However, prior to this, the issue of the limiting values of the Reynolds number is addressed.

4. Limiting values of Reynolds number Bearing in mind the mean experimental error of 5% in the experimental determination of f at low Reynolds num-

and K 0:6628 1:458exp0:028175fRel

8d

56

R.P. Chhabra et al. / Powder Technology 129 (2003) 5358

Table 2 Limiting values of Rel marking the end of viscous region k 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Rel 0.027 0.040 0.050 0.083 0.18 0.52 2.10 8.40 25.17

number of data points for each source have been delineated using this criteria. Table 1 also gives the distribution of data points in each flow regime.

5. Prediction of wall factor The entire data population was thus subdivided into three categories depending upon the value of the Reynolds number using the criterion outlined above. As mentioned earlier, undoubtedly, the excellent prediction capability of Eq. (4) in creeping flow [1 3] and of Eq. (5) in turbulent flow [3,11,12,25] have been demonstrated beyond reproach. Suffice it to add here that Eq. (4) predicts most of the literature data with a mean deviation of 6 7%, while the corresponding value for Eq. (5) in the inertial region is of the order of 8 9%, with a notable exception of the data reported in Ref. [25]. For this set of results, f f 1 for all values of k. The reasons for this unusual behaviour are not immediately obvious. Both these figures are well within the limits of experimental uncertainty, and these are as such regarded to be satisfactory and acceptable. Attention is now turned to the prediction of the value of f in the intermediate Reynolds number regime. Depending upon the value of k and Re, the efficacy of all the four methods, namely, Eq. (7), Eqs. (8a) (8d), Eqs. (9a) and (9b), Eq. (10) and Eq. (12) presented in the preceding section has been tested for the estimation of the value of the wall factor in the intermediate Reynolds number regime.

bers, the critical Reynolds number is defined here as the value at which the value of the wall factor is 5% greater than that estimated from Eq. (4). Based on the application of this criterion to the experimental results available in the literature and listed in Table 1, the resulting limiting values of the Reynolds number are summarised in Table 2; it is clearly seen that the so-called viscous region persists up to higher and higher values of Rel as the value of k increases. It is, however, appropriate to mention here that Eq. (7) yields the limiting values of Rel which decrease with the increasing value of k. Clearly, this trend is at variance with the experimental results available in the literature. However, the reasons for this discrepancy are not immediately obvious. On the other hand, in the fully turbulent region, the experimental values of f often entail an error of the order of 10%. Thus, the critical value of the Reynolds number beyond which f can be assumed to be a function of k only, approximated here by Eq. (5), was estimated from the available experimental results as the value of Re at which the value of f is within 10% of that predicted by Eq. (5). Table 3 summarises the values of the Reynolds number obtained by using this procedure [12]. Here also, the critical value of Re is seen to increase with the increasing value of k. Unfortunately, lack of data for k>0.85 precludes the evaluation of the critical value of Re for k = 0.90 in this case. Further detailed examination of the behaviour of results in the vicinity of transition zone [8,9,12] revealed that no discernable trends are present with respect to the value of k. Thus, the values summarised in Tables 2 and 3 can now be used to delineate the flow regime for the evaluation of the wall factor for a sphere, and the total
Table 3 Limiting values of Re marking the onset of fully turbulent region [12] k 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.85 Re 60 110 200 500 2000 6700 104 1.25 104 1.5 104

Fig. 1. Comparison between predictions of Eq. (7) and experimental results of wall factor, f.

R.P. Chhabra et al. / Powder Technology 129 (2003) 5358

57

Fig. 2. Comparison between predictions of Eqs. (9a) and (9b) and experimental results of wall factor, f.

Fig. 4. Comparison between predictions of Eqs. (10) and (12) and experimental results of wall factor, f.

While Figs. 1 4 show the parity plots in terms of the calculated versus experimental values of f, Table 4 provides an overall summary of the results in terms of average and maximum deviations together with the number of data

points. Since each method tested here is applicable over different ranges of conditions, the number of data points tested varies from one method to another. An examination of the results shown in Table 4 (and in Fig. 2) clearly shows that Eqs. (9a) and (9b) due to DiFelice [22] yield overall minimum average deviations between the predictions and the experimental values, albeit the maximum deviation of 158% is rather discomforting. In addition, there does not appear to be any discernable trends present in these figures except that the results of Uhlherr and Chhabra [9] seem to be generally under-predicted. However, a detailed examination of the results of Uhlherr and Chhabra [9] showed that there are about 23 data points, all relating to k>0.87, account for such large deviations. It is readily conceded that for such large values of k coupled with rather high values of the Reynolds number, spheres have a propensity to follow a zigzag trajectory and are known to undergo rotation also during the course of settling. Besides, as the value of k and/ or of the Reynolds number increases, the accuracy of experimental results deteriorates rather rapidly [9]. If these points are excluded from comparison, though the average
Table 4 Summary of results Method % Error Average Eq. (10) and Eqs. (11a) (11c) Eqs. (10) and (12) with p = 2.2 Eq. (7) and Eqs. (8a) (8d) Eqs. (9a) and (9b) 18.74 25.88 27.10 12.31 Maximum 405 575 518 158 Number of data points 387 387 467 604

Fig. 3. Comparison between predictions of Eq. (10) and Eqs. (11a) (11c) and experimental results of wall factor, f.

58

R.P. Chhabra et al. / Powder Technology 129 (2003) 5358

error drops only marginally to 12%, the maximum deviation reduces to 29%. However, since these 23 data points are outside the range of applicability of the other equations, the exclusion of these data points from the comparisons does not alter the mean and maximum errors associated with Eq. (7), Eq. (10) and Eqs. (11a) (11c) and Eqs. (10) and (12). Furthermore, in assessing these comparisons, it must be borne in mind that even a small degree of nonverticality of tubes contributes appreciably to the error in the falling velocity. Extensive attempts were made to improve the predictability both by altering the values of the fitted constants in the existing methods and by using new functional forms. However, it soon became evident that it was possible to improve the predictions of a specific set of data at the expense of others, but without resulting in any significant overall improvement. Besides, the mean error of 12% associated with Eqs. (9a) and (9b) is comparable to the experimental accuracy of the results. Finally, before concluding this section, an observation about the efficacy of the only theoretical expression, Eq. (7) and Eqs. (8a) (8d), is also in order. This method seems to work well for small values of k and/or the Reynolds number. However, the greater the value of k, the lower was the value of the Reynolds number at which the predictions and experimental data began to diverge appreciably. For instance, for k V 0.3, this method yielded satisfactory predictions up to Rel f 100, whereas this value dropped to 10 for k = 0.4, and for k z 0.5, the match between these predictions and experimental results was poor for all values of Reynolds number, as can be seen in Fig. 1. In summary, Eqs. (4) and (5) must be used to estimate the value of f in the so-called viscous and fully turbulent regions, respectively, whereas the use of the empirical expression given by Eqs. (9a) and (9b) is recommended in the intermediate range. The values of the Reynolds number listed in Tables 2 and 3 can be used to delineate the flow regime.

gives the best prediction of the wall factor in the low Reynolds number regime, whereas under fully turbulent conditions, the use of Eq. (5), due to Newton [20], is recommended. In the intermediate regime, the empirical formulation of DiFelice [22], Eqs. (9a) and (9b), correlates most of the literature data with an overall minimum error of about 12.5%, albeit deviations increase steeply with the increasing values of k and/or the Reynolds number. For k>0.88 or so, only limited data is available, and thus, extreme caution must be exercised in using any of the available correlation under these conditions.

Acknowledgements Helpful discussions with Professor DiFelice, particularly in interpreting the theoretical results leading to Eq. (7), are gratefully acknowledged.

References
[1] R. Clift, J. Grace, M.E. Weber, Bubbles, Drops and Particles, Academic Press, New York, 1978. [2] R.P. Chhabra, Bubbles, Drops and Particles in Non-Newtonian Fluids, CRC, Boca Raton, FL, 1993. [3] R.P. Chhabra, in: D. DeKee, R.P. Chhabra (Eds.), Transport Processes in Bubbles, Drops and Particles, vol. 2, Taylor and Francis, New York, 2002, pp. 318 339. [4] H.S. Munroe, Trans. AIMME 17 (1888-89) 637. [5] H. Faxen, Ask. Mat. Arstron. Fys. 17 (1923) 1. [6] W.L. Haberman, R.M. Sayre, David Taylor Model Basin Report No.1143, Dept. of Navy, Washington, DC, 1958. [7] V. Fidleris, R.L. Whitmore, Br. J. Appl. Phys. 12 (1961) 490. [8] R.P. Chhabra, P.H.T. Uhlherr, Chem. Eng. Commun. 5 (1980) 115. [9] P.H.T. Uhlherr, R.P. Chhabra, Can. J. Chem. Eng. 73 (1995) 918. [10] M. Iwaoka, T. Ishii, Experimental wall correction factors of single solid spheres in circular cylinders, J. Chem. Eng. Jpn. 12 (1979) 239. [11] R. DiFelice, P.U. Foscolo, L.G. Gibilaro, Chem. Eng. Sci. 50 (1995) 3005. [12] R.P. Chhabra, P.H.T. Uhlherr, J.F. Richardson, Chem. Eng. Sci. 51 (1996) 4531. [13] H.M. Lee, MS Thesis, Univ. of Iowa, Ames (1947). [14] M.B. McPherson, MS Thesis, Univ of Iowa, Ames (1947). [15] S.M. Engez, MS Thesis, Univ. of Iowa, Ames (1948). [16] A.R. Khan, J.F. Richardson, Chem. Eng. Commun. 62 (1987) 135. [17] P.L. Paine, P. Scherr, Biophys. J. 15 (1975) 1087. [18] W.R. Bowen, A.O. Sharif, J. Colloid Interface Sci. 168 (1994) 414. [19] J.J.L. Higdon, G.P. Muldowney, J. Fluid Mech. 298 (1995) 193. [20] I. Newton, Principia. Lib. II, Prop. XXXIX, Theorem XXXI, Cambridge University Press, Cambridge (1987). [21] R.M. Wham, O.A. Basaran, C.H. Byers, Ind. Eng. Chem. Res. 35 (1996) 864. [22] R. DiFelice, Int. J. Multiph. Flow 22 (1996) 527. [23] R. Kehlenbeck, R. DiFelice, Chem. Eng. Technol. 21 (1999) 303. [24] V. Fidleris, PhD Thesis, Univ. of Nottingham, U.K. (1958). [25] A. Bougas, M. Stamatoudis, Chem. Eng. Technol. 16 (1993) 314.

6. Concluding remarks Based on a detailed analysis of the literature data on wall effects for a sphere falling in incompressible Newtonian media in cylindrical tubes, the limiting values of the Reynolds number denoting the end of the so-called viscous and the onset of the fully inertial or turbulent regions have been suggested. These values, in turn, have been used to classify the bulk of the literature data on wall effects. In addition, the relative performance of some of the prominent analytical and/or empirical expressions has been assessed. Based on extensive comparisons, it is suggested that the analytical expression of Haberman and Sayre [6], Eq. (4),

Vous aimerez peut-être aussi