EXPERIMENTAL WALL CORRECTION FACTORS OF SINGLE SOLID SPHERES IN CIRCULAR CYLINDERS MlTSUGU IWAOKA AND TSUTOMU ISHII Department of Chemical Engineering, YokohamaNational University, Yokohama240 Introduction Concerning the wall effect on the slow motion of single solid spheres in circular cylinders some studies1'4"9'13>15) and theoretically. treated this methods. have been made both experimentally Most problem theoretical with The experimental a series onesl!4'7~9'13>15) of expansion ones5>6'14il7) were based on measurements of the settling velocities of single solid spheres in circular cylinders. However, it cannot be said that there are adequate, reliable data for the wall correction factors of single solid spheres in circular cylinders. In the flow fields of packed beds, particulate fluidization and hindered settling, particles interfere each other and as a result the drag coefficient with of a particle in a swarm of particles is larger than that of a single particle in an infinite medium, as the fluid velocity on the surfaces of all particles is zero and therefore the resistance force becomes larger for fluid to pass through such a swarm. The relationship between the drag coefficient of a particle in a swarm and that of a single particle in an infinite mediumhas been expressed by the following void function defined by Ito11>12), on which previous theoretical and experi- mental work is summarized by Ishii10). CDX=CDs F{e) (1) Also, in the case of a single particle settling in a finite medium surrounded by a solid wall on which the fluid velocity is zero, the particle interferes with the solid wall. The drag coefficient of a particle in a finite mediumis larger than that of a particle in free settling, if the particle Reynolds number is equal in both cases. This is because the resistance force becomes larger for a particle to pass through such a narrow space surrounded by the solid wall. This relationship can be expressed in a similar way by the CDw = CDsF(r) * (2) Here, F(r) is called the wall correction factor. Received June 17, 1978. be addressed to T. Ishii. VOL 12 NO. 3 1979 Correspondence circular cylinders as a basic study of the flow fields of packed beds, particulate fluidization and hindered settling. This study resulted in the establishment of the precise design for a falling ball viscometer, as well as a basic study on packed beds, particulate fluidi- zation and hindered settling. 1. Experimental Four glass tubes of about 1 cm I.D. and 100cm in length were used as experimental apparatus. A thermometer was inserted into one of the tubes, containing the same liquid as the others, for measuring the temperature under the same conditions as the actual experiment in the other tubes. This experi- ment was done in an air-conditioned room, in which the room temperature was kept constant at about 20°C. The tubes for the actual experiment were 0.900, 0.990 and 1.00 cm I.D. and were supported vertically by making the tubes parallel weight. The entrance to a thread and end region suspending are necessary a due to the unsteady motion of solid spheres influenced by the upper and lower surface, respectively. Thus, the distance of 20cm was taken for both entrance and end region of this experiment and seems to be adequate, based on fairly many previous studies. Therefore, the section for measuring the time in which the spheres fall through was 60 cm in the middk of the circular cylinders. r.;à"å -. - Millet jelly aqueous solution was used as a liquid for this experiment. It is well-known that this is a highly viscous Newtonian fluids. Four millet jelly aqueous solutions of different concentrations, designated as A, C, D and E, were prepared by dissolv- following equation. 051. Thus the void function and wall correction factor are proved to have similar properties. The purpose of this workwas to measure accurately the wall correction factor of single solid particles in concerning this article should M. Iwaoka is now at Japan Steel Works, Muroran ing given amounts of millet jelly into warmwater. The viscosities of these solutions were about 2800, 330, 760 and 2000 centipoises respectively, at 20°C. The viscosities were determined with a Hoeppler falling ball viscometer at constant temperatures of ap239 Table 1 Previous work 1. 2. 3. 4. 5. Landenburg13) Faxen4> Faxen4) Lee14> 4. 1 Bohlin1) 7. Haberman and Sayre8> Suga17> 10. Francis6) ll. Fidleris l +9/4*+(9A*)2 2 7r/8)(l -r)-5/2 F(r)=(l/l -2.105r+2.087*3) F(r)=(l -0.75857r5)/(l F(r)=(l -r)-2-25 F(r)={(l-0.475r)/(l-r)}4 and Whitmore5) r ^0.3 0.7^*^ r^0.3 1 solutions. Bearing balls of diameters from 1 to 9mmwere used as solid spheres for this experiment. 2. Results and Discussion The particle Reynolds number Repw, in which the settling velocity uw of the sphere influence by the wall effect was used as a representative velocity, is less than 0.0762. The particle Reynolds number RePF, in which the settling velocity infinite medium calculated ?=0.3 -2.1050r+2.0865*3-1.7068*5+0.72603r6) rko.5 r^0.832 with glucose for rotating spheres with glycerol r^0.970 Presented graphs proximately 20, 25, 30, 35 and 40°C. Distilled water and a standard liquid for correcting viscometers, prescribed by JIS Z8809, were used as 240 *^0.3 *^0.3 Exact solution of algebraic equations More exact solution of more than 8 algebraic equations. Both solutions are almost the same curve in the figure. F(*)=l/(l -r)2-1/w-*25 *^0.333 Fig. 1 Previous wall correction factors of single solid spheres in circular cylinders standard *6) F(*)=l/(l-2.10443*+2.08877r3-6.94813*5-1.372*6+3.87r8-4.19r10) 8.1. Haberman7) 8.2. Paine and Scherr15) 9. r^O.l rWl F(r)=1/(1 -2.104*+2.09*3-0.95*5) F(*)=(l-V4*)/(l-V+5/2*3-9/4*5+ 4.2 F(*)= 4.3 F(?)=(3V and Byrne9) Happel 6. F(r)=l+2Ar F(r)=1+2.1* uF of the sphere in an by Stokes drag formula Fig. 2 Experimental wall correction factors of single solid spheres in circular cylinders was used, is less than 0.748 except that the same numbers were 3.02 and 2.53 in the case ofDp=0.9 cm. In such a creeping-flow region the followingrelationship holds for the wall correction factor, which is calculated from this equation. F(r) = uF/uw (3) All the previous studies are summarized in Table 1 and plotted in Fig. 1, where the correlating equations are considerably convergent in the small r ranges (solid lines), but they were extrapolated beyond the applicable ranges (dotted lines). The plotted data show large scatter as r increases and it is not known which correlating JOURNAL equation is best for estimating OF CHEMICAL ENGINEERING OF the JAPAN direction opposite to the direction for rolling along the wall effect. Figure 2 shows a semi-log plot of the wall correction factors of this experiment and dimensionless radius of the sphere. The data of glucose ofFrancis6} and those of Fidleris and Whitmore5) and McNown, Lee, McPhersonand Engez14) are also plotted in the same figure. The data of glycerol of Francis were omitted, as they concerned rotating spheres compared with the other theories2'3). and should be The data of this experiment agreed very well with the exact theory of Haberman7), and Paine and Scherr15). But the data of McNown,Lee, McPherson and Engez14), Fidleris and White5\ and Francis6) are lower for wide ranges of rjR. Furthermore, although up to r=0.8 solid spheres were never rotating, in the case of r approximating 0.9 for this experiment the solid spheres were always eccentrically rotating as they did when the cylinder was tilted. The imperfect roundness of the cross-section, imperfect verticality and curvature of the circular cylinder cause a solid sphere to rotate in part in the case of very small clearance. But even though the necessary precautions were taken, the occurrence of rotation was unavoidable. According to McNown, Lee, McPherson and Engez14), at moderately low values of the Reynolds number in the creeping-flow region the sphere settled without rotation in a position practically concentric with the cylinder, but in this nearer side of the cylinder wall. The settling velocity increases monotonically with lateral position until about e=0.98, in which position the sphere settles approximately 2.1 times faster than for e=0. For e>0.98 the rate of settling decreases, as does the angular velocity. Their perturbation solution further predicts that at some lateral position e, close to unity, the direction of rotation changes. Although the settling velocity undergoes a two-fold change with lateral position remains The spheres in the range 0<e<0.98, the pressure drop essentially constant over the same range. data of this experiment except the rotating was correlated by the following equation using the least-squares method. F(r)=l/(1.000000+1.137853p-0.2644559 +0.8474067x 10V+0.1128310X -0.1003350x +0.1922832X +0.5989260X l0V5+0.1252298 10V-0.6372137-X 10V-0.1967150 x l0-V2 10V4 * x lOV6 10V8 X 10V° (4) In conclusion it turns out that the exact theory of Haberman7) and the more exact solution of Paine and Scherr15) can be applied to falling ball viscosity meas- urements except for very small clearances between a solid sphere and circular cylinder, and the fluid re- sistance force for a macromoleculein restricted diffusion can be precisely estimated. In the case of very small clearance more precise measurements of rotation speed, rotation direction, lateral position of a sphere as well as settling velocity require the use of still and 8-mm low-speed cameras in future studies. experiment there was always with rotation in a position eccentric with the number range. cylinder in the same Reynolds Nomenclat ure Christopherson and Dowson3) applied an elementary lubrication theory to the settling phenomena of a solid sphere in a circular cylinder in the case of small clearance and found that the settling velocity of a solid sphere with rotation in a particular eccentric position becomes largest and therefore the drag coefficient minisuch an eccentric settling with rotation was predicted as stable from the viewpoint of minimumviscous dissipation. This theory was based only on the equation of motion offluid, but to obtain a conclusion from the viewpoint of minimumviscous dissipation, the equation of energy of fluid, must be considered from the viewpoint of the minimumentropy production rate. The direction of rotation in the eccentric position in this experiment is opposite to that reported by christopherson and Dowson3). Bungay and Brenner2} have established a more precise lubrication theory, using perturbation method, for such a settling problem. In the case of sedimentation of a sphere in a vertical cylinder a concentric sphere (e=0) settles without rotation. If displaced to an intermediate position, rapidly. Simultaneously, VOL. 12 NO. 3 1979 the sphere descends more it undergoes rotation in a e mum.Therefore, perpendicular distances from sphere center F*r) F(e) R uw to cylinders axis drag coefficient of a single sphere in an infinite medium drag coefficient of a single sphere influenced by the wall effect drag coefficient of a particle in a swarm diameter of a solid sphere=2r eccentricity bj(R-r) wall correction factor void function radius of a circular cylinder radius of a solid sphere = dimensionless radius of a solid sphere=r/R [-] [-] [cm] [-] [-] [-] [cm] [cm] [-] settling velocity of single solid sphere in an infinite medium (expressed by Stokes [cm/sec] drag formula in this case) settling velocity of a single solid sphere [cm/sec] influenced by the wall effect = viscosity of solution = density of solution [c.p. or poise] [g/cm3 ] 241 [g/cm3] = density of a solid sphere Literature Cited 1) Bohlin, T.: Trans. Roy. Inst. Technol, Stockholm, No. 155 (1960). 2) Bungay, P. M. and J. Brenner: Int. J. Multiphase Flow, 1, 25 (1973). 3) Christopherson, A251, 550 (1959). D. G. and D. Dowson: Proc. Roy. Soc, 4) Faxen, H.: Ark. Mat. Astron. Fys., 17, No. 27 (1923). 5) Fidleris, V. and R. L. Whitmore: Brit. J. Appl. Phys., 12, 490 (1961). 6) Francis, A. W.: Physics, 4, 403 (1933). 7) Haberman, W. L.: Ph.D. Thesis, University (1956). Basin Dept. Report (1958). No. 1143, Washington, D.C., U.S. Navy Happel, J. and BJ. Byrne: Ind. Eng. Chem., 46, 1181 (1954). Ishii, T. : Kagaku Kogaku, 29, 915 (1965). Ito, S. : "Kagaku-Kogaku-Nyumon", Kagaku-Kogyo-Sha, Tokyo (1957). Ito, S.: Kagaku Kogaku, 26, 1108 (1962). Landenburg, R. : Ann. Phys., 23, 447 (1907). McNown, J. S., M. J. Lee, M.B. McPherson Engez: Proc. p. 17 (1948). 7th Intern. Cong. Appl. and S. M. Mech., London, Paine, P. L. and P. Sherr: Biophysical J., 15, 1087 (1948). Sato, T., I. Taniyama and S. Shimokawa: Kagaku Kogaku, of Maryland 30, 34 (1966). Suga, S.: Bull. Inst. 146 8) Haberman, W. L. and R. M. Sayre: David Taylor Model (1931). Phys. and Chem. Research, Tokyo, 10, NATURAL CONVECTION HEAT TRANSFER FROM HORIZONTAL CYLINDERS Jagdish CHANDand Dharam VIR Department of Chemical Engineering and Technology, Panjab University, ponent continuously Introduction The problem of natural convection from horizontal cylinders was treated8>23'25'31) as one of steady-state conduction in a stationary gas film. However, there is ample evidence that heat transfer occurs by convection streams within the so-called film itself. Analytical treatment is usually based on the set of equations representing the conservation of mass, momentumand energy in a fluid moving under the influence of body force. 0.25 in the laminar range and 1/3 in the range. A summary of correlations of this type is given in Table 1. However, experimental data on horizontal cylinders collected by Fishenden continuously decreases as (Pr à" Gr) is increased beyond unity. For vertical plates, Ostrach29) numerically solved the equations for 0.01<Pr< 1000 and calculated profiles for laminar flow to obtain Nu=a(Pr-Gr)0 25. The computer solution of Ostrach was shown by Le Fevre24) to fit the empirical equation a4-Pr/(2.43478+4.884Pr°-5+4.95283Pr) Further calculations of the same type have been re- lationship between log Nu and log (Prà" Gr) is essentially non-linear and the experimental slope maydeviate from 0.25, particularly when Prà"Gr is large. Table 2 shows the values of the coefficients and exponents at different levels of (Pr-Gr). It is evident that the exReceived November 2, 1978. should be addressed to J. Chand. Correspondence concerning this Qualitatively a similar situation showing that a= F(Pr) is expected to occur for horizontal cylinders. This affects the linearity of the relationship between log Nu and log (Prà"Gr). Another reason for the nonlinear relationship is the presence of a temperature ratio term in the relationship by Hermann13). Nu=F(Pr, and Saunders9) and shown in Fig. 1 suggests that the re- 242 increases and that the coefficient ported10>27'34'35). The principle of similitude5)6'20)28) and suitable simplifying assumptions in the boundary layer theory produce correlating equations of the type Nu=F(Pr, Gr) and in particular Nu=aPrhGrc with b and c usually equalling turbulent Chandigarh 160014 India article Gr,(Tw - T^)/ TOQ) Hymanet al.16) correlated natural convection data with 14 different liquids including liquid metals by the equation Nu=0.53 [{Pr1(0.952+Pr)}(Pr Hausen12) recommended à" Gr)]0 the following 25 empirical equation : Nu=0.U(Pr'Gr)1/3+(Pr'Gry \ lO~7<Pr-Gr< 1012 JOURNAL OF CHEMICAL ENGINEERNG OF JAPAN