Y. S.

Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, NF, A1B 3X5, Canada

Pressure Drop in Laminar Developing Flow in Noncircular Ducts: A Scaling and Modeling Approach
A detailed review and analysis of the hydrodynamic characteristics of laminar developing and fully developed flows in noncircular ducts is presented. New models are proposed, which simplify the prediction of the friction factor–Reynolds product f Re for developing and fully developed flows in most noncircular duct geometries found in heat exchanger applications. By means of scaling analysis it is shown that complete problem may be easily analyzed by combining the asymptotic results for the short and long ducts. Through the introduction of a new characteristic length scale, the square root of cross-sectional area, the effect of duct shape has been minimized. The new model has an accuracy of 10% or better for most common duct shapes when nominal aspect ratios are used, and 3% or better when effective aspect ratios are used. Both singly and doubly connected ducts are considered. DOI: 10.1115/1.4000377 Keywords: laminar flow, noncircular ducts, hydrodynamic entrance length, developing flow, modeling, pressure drop

M. M. Yovanovich
Department of Mechanical and Mechatronics Engineering, University of Waterloo, Waterloo, ON, N2L 3G1, Canada



Laminar flow fluid friction and heat transfer in noncircular ducts occur quite frequently in low Reynolds number flow heat exchangers such as automotive coolers, cold plates, and microchannel heat sinks. It is now also occurring more frequently in a host of other energy conversion and reclamation devices, as a result of the miniaturization of modern technologies. While traditional approaches have relied heavily on the use of tabulated and/or graphical data, the ability to design thermal systems using robust models is much more desirable in the age of computer simulations and computer assisted design. Most modern fluid dynamics and heat transfer texts rarely present correlations or models for more complex geometries, which appear in many engineering systems. Rather, a subset of data for miscellaneous geometries is usually presented after detailed discussion and analysis of simple geometries such as the circular duct and parallel plate channel. In the present paper, the hydrodynamic problem is considered in detail, and a new and much simpler model is developed for predicting the friction factor–Reynolds number product for developing laminar flow in noncircular ducts. Laminar fully developed fluid flow in noncircular ducts of constant cross-sectional area results when the duct length L is sufficiently greater than the hydrodynamic entrance length Lh, i.e., L Lh, or when the characteristic transversal scale is sufficiently small to ensure a very small Reynolds number. Under these conditions the flow through most of the duct or channel may be considered fully developed. However, in many engineering systems such as compact heat exchangers and microcoolers used in electronics packaging, while the characteristic dimension of the flow channel is small enough to give rise to laminar flow conditions, the flow length is generally not sufficiently large enough to give rise to fully developed flow, i.e., L Lh or L Lh, and developing
Contributed by the Fluids Engineering Division of ASME for publication in the JOURNAL OF FLUIDS ENGINEERING. Manuscript received March 19, 2009; final manuscript received September 23, 2009; published online October 27, 2009. Editor: Joseph Katz.

flow prevails over most of the duct length. In these situations, a model capable of predicting the hydrodynamic characteristic, usually denoted as f Re, the friction factor–Reynolds number product as a function of dimensionless duct length, is required. Since the pressure drop in a developing flow is due to both wall shear and fluid acceleration, some references choose to denote it as the apparent friction factor–Reynolds number product f app Re, in order to distinguish it from the fully developed flow value f Re. However, since dimensionless pressure drop or f app Re in the entrance region transitions smoothly to f Re in fully developed flow, this distinction is dropped in favor of just f Re, which is implied to vary with dimensionless duct length, i.e., f Re L+ , where L+ is the dimensionless duct length to be defined shortly.


Literature Review

A review of literature reveals that only two significant attempts at developing a general model have been undertaken. These are the work of Shah 1 and Yilmaz 2 . Both of these models are based on the earlier work of Bender 3 . Bender 3 combined the asymptotic result for a “short” duct 4 , with the result for the “long” duct, to provide a model that is valid over the entire length of a circular duct. The demarcation between short and long being the duct length relative to the hydrodynamic entrance length as shown in Fig. 1, i.e., 3.44 f ReDh = L+ ,
fd +

L+ K , L+ 4L+ L/Dh ReDh

0.001 1 0.1

f Re where

L+ =


is the dimensionless duct length. As such, this formulation requires the use of the concept of the incremental pressure drop factor K . Shah 1 later extended the model of Bender 3 to predict results for the equilateral triangle, NOVEMBER 2009, Vol. 131 / 111105-1

Journal of Fluids Engineering

Copyright © 2009 by ASME

Downloaded 27 Oct 2009 to Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

most available data are a result of solution to a much simpler set of equations. This model is more general than that of Shah 1 but is also quite complex since a base correlation has been augmented with three additional correlations. the model of Yilmaz 2 is accurate over the entire range of the entrance and fully developed regions for many duct cross sections. In this approach the incremental pressure drop factor K is required in the long duct solution see Eq. and friction. V = 0. the boundedness condition along the duct axis.153. It is apparent from the available data that smooth transition occurs from the entrance region to that of fully developed flow. 2 Common duct shapes Transactions of the ASME Downloaded 27 Oct 2009 to 134. 4 accounts for the increase in momentum of the accelerating core and the wall shear. the use of the term K in a hydrodynamic entrance model such as that proposed by Bender 3 is redundant. integral methods. It was formulated as a nonlinear superposition of two asymptotes. equations subject to the no slip condition at the duct wall. For two dimensional cross sections. which appears in the Shah 1 model. and theoretical approaches. Simultaneous solution to the continuity. Several of the duct shapes of interest are shown in Fig. whose solution varies with aspect ratio. Eq. K . 6 Geometry Rectangle Shape 2b 2a P/ A 2( + 1) Ellipse b a 4 E( ') r* = ro N = # sides Annulus ri ri ro 2 (1+ r ) * (1 r*2 ) 1/ 2 Polygons b 3 Governing Equations Isosceles Triangle 2b 2 2 N tanh 2 cos N The governing equations for steady incompressible flow in the hydrodynamic entrance region in a noncircular duct or channel are ·V=0 and 111105-2 / Vol. this is achieved at the cost of simplicity. and C. Later. The momentum equation represents a balance of three forces: inertia. however. i. Unfortunately. 1 . thus limiting interpolation for geometries such as the rectangular duct and circular annulus. The primary drawback of the simple model proposed by Shah 1 is the requirement of tabulated coefficients and parameters for each geometry. and by tabulating coefficients for each particular geometry for a number of channel and annulus aspect ratios using f ReDh = 3.. pressure. As a result of the complex correlating equations for K developed by Yilmaz 2 . Despite its complexity. variational methods. scaling analysis is used to show the appropriate form of the solution for both short and long ducts.160. the model introduces the incremental pressure drop. the incremental pressure drop K . However. which is not actually required. 4. 4 . which contains the true fully developed asymptote. see http://www. 4 . However. 5 . Solutions to Eqs. the z-momentum equation. The two models discussed above represent the current state of the art for internal flow problems. i.e. Since the solution obtained by Shapiro et al. In both cases.org/terms/Terms_Use. only one of these asymptotes is a true asymptote. Rather than tabulating coefficients. In the case of cross sections that can be characterized by a single variable. In Sec. this form of correlation is more complicated than need be for this problem. and momentum.cfm . the models are based on the proposed form put forth by Bender 3 . Both models are based on the combination of the short duct and long duct solutions using the correlating method of Bender 3 . Yilmaz 2 developed a complex correlation scheme for the fully developed friction factor f Re fd. Additional discussion on special issues pertaining to the solution of this problem is discussed in Ref. which predicts that all of the available data are now developed. Redistribution subject to ASME license or copyright.44/ L+ 1 + C/ L+ 2 3 Yilmaz 2 proposed a more general model of Shah 1 ..e. 1 Hydrodynamic entrance problem the circular annulus.asme. 4 and 5 have been obtained using a host of methods including finite difference methods.44 L+ + f Re fd + V· V=− p+ 2 V 5 K /4L+ − 3. Shah 1 achieved this by generalizing the form of the model of Bender 3 . i. The model presented in this paper does not require this parameter and is significantly simpler in composition. In the case of the model developed by Yilmaz 2 . which result from boundary layer type assumptions 4 . an additional closure equation is required for relating transverse velocity components.27. 4 Scale Analysis We now examine the momentum equation and consider the various force balances implied under particular flow conditions. is required to characterize the flow. which ultimately require some form of numerical method. As such.x w z δ Lh Developing Flow Region Fully Developed Flow Region Fig. the rectangular duct.. 2. these yield the continuity equation and momentum equation in the direction of the flow. interpolation is no longer a problem. and parallel plate channel geometries. the simple physical behavior of the hydrodynamic entrance problem is lost.e. however. Eq. NOVEMBER 2009 4 = h 2b + 2 tan tan Fig. f Re fd. and a fitting coefficient C. A simple modeling approach is presented. The other asymptote represents a transitional one. 131. and a constant initial velocity. V . V = Uk. asymptotic analysis is used to develop a new model.

which gives us freedom to choose L. The major distinction. which results from the force balance. we will examine the various force balances implied by Eq. 12 .e. considering the relationship for the wall shear = V 21 and the friction factor. both experience boundary layer growth. Thus we may write the following approximate relation using the characteristics of the flow and the geometry: p L U L2 8 4. 1 2 2 U .org/terms/Terms_Use. that hydrodynamically developing flows should be presented as a function of L+ rather than L+. Eqs. The force balance within the boundary layer is governed strictly by inertia and friction forces 16 This balance yields U2 L where 1 L ReL U 2 17 where L represents a characteristic transversal length scale of the duct cross section. is not the most appropriate choice. since the effective wall shear is inversely proportional to and hence a function of L+ as shown in Eq. 14 . but constant cross section. 5 in the two regions divided by L h. Each is examined below using the method of scale analysis advocated by Bejan 5 . 1 as a starting point. due to a favorable pressure gradient. Equation 12 represents a dimensionless mean wall shear. L š Lh.153. Fully developed laminar flow in a duct of arbitrary. Thus. Rearranging the above expression gives p UL L2 9 is the boundary layer thickness see Fig. see http://www. as we shall see is the rate at which the boundary layer grows within a confined channel. we may introduce the definition of the friction factor.We now consider three separate force balances. which is omitted in the scaling analysis. we obtain f U 2 U/ U2 ReL UL 1 ReL 22 Equation 12 is often denoted as the Poiseuille number Po in modern literature. except that the velocity at the edge of the boundary layer is no longer constant in the duct flow. 10 and 14 : L A P Dh 4 15 We shall see later that this length scale. 11 . 4. Finally. 1 . In the entrance region near the duct inlet two regions must be considered. L ™ Lh. is much slower than a plate in streaming flow.asme.cfm . Introducing the transversal length scale L we obtain L L L ReL L L ReL U L+ 19 Next.160. when L = Dh. such that f ReL = 2 PoL. The shear stress may be approximated by = V U L 10 Although the flow in a confined channel is markedly different from that of a plate in streaming flow. The two regions are now examined beginning with the flow within the boundary layer. i.. The flow within the boundary layer is very similar to laminar boundary layer development over a flat plate. in the inviscid core. The factor of 2 results from the use of the kinetic or dynamic pressure in Eq. and defining the product of friction factor and Reynolds number yield the following expression for the entrance region: f ReL or f ReL = C2 L+ 24 O1 L+ 23 and yields the following relationship when combined with the two scaling laws. The velocity scales according to the area mean value U. Using Fig.27. but this by no means implies necessity of its use in Eq. As a length scale it is fundamentally important in the relationship of Eq. Redistribution subject to ASME license or copyright. although convenient. and the axial length scales according to L. 131 / 111105-3 Downloaded 27 Oct 2009 to 134. 18 Next. defined as the dimensionless wall shear f U/L U 2 2 where L+ is the dimensionless duct length defined by L+ = 20 U 1 ReL 11 The above expression may be written such that the following relationship exists for all cross-sectional geometries: f ReL or f ReL = C1 13 L U O1 12 Next. One is the inviscid core and the other the viscous boundary layer. Vol. we examine the shear stress at the wall. 24 . the momentum equation represents a balance of inertia and pressure forces 25 which scales according to U2 p 26 NOVEMBER 2009. which is inversely proportional to the Reynolds number. It is in this definition that the appropriate choice of a characteristic length scale must be carefully considered. The constant C1 has been found to vary for most geometries in the range 12 C1 24. Finally. Journal of Fluids Engineering Equation 19 illustrates an important point highlighted by Bejan 5 . is governed by the Poisson equation 7 which represents a balance between the friction and pressure forces.1 Long Duct Asymptote. a control volume force balance gives pA = L P 14 Rewriting the above expression in terms of the length scale L.2 Short Duct Asymptote.

cfm . If we consider a simple channel formed by two plates separated by 2b. by considering other problems in mathematical physics for which the Poisson equation applies.160. In summary. 19 ..12 . It typically has an order of magnitude 0. However. perimeter P. semi-axis lengths.. the hydraulic diameter has been chosen. A. when their thickness is of order of the half plate separation = or Lh 0. defined by Eq.. Duan and Muzychka 13 and Muzychka and Edge 14 showed that the square root of flow area is also more appropriate for nondimensionalizing gaseous slip flows and non-Newtonian flows. by considering an equality between the two asymptotic limits given in Eq. The result of these two effects is that a mismatch occurs between the true hydraulic equivalent duct and the hydraulic equivalent duct determined by the definition 4A / P.153. i.e. This estimate is much shorter than the actual length in a channel flow. 15 . The use of the hydraulic diameter is often considered the most convenient as a result of the relationship defined by Eq. the authors addressed this issue using dimensional analysis. In a laminar flow the velocity gradient varies over most of the cross-sectional area and flow does not penetrate as deep into re-entrant corners. L.011Dh ReDh 31 which is approximately four times greater. and length L. AL P 1/3 We use the much greater and much less than notation. which minimizes the effect of duct shape on the numerical value of the dimensionless friction factor–Reynolds number product. see http://www. the constant in Eq.328 and C2 = 0.Since the boundary layer in the developing region will be thinner under a favorable pressure gradient in the core. However. 21 . However. Traditionally.72 for a local friction factor. L = 4A / P. given the inverse dependency of f on . the dimensionless grouping f ReDh takes on values that vary between 12 and 24 for most common duct shapes.0025Dh ReDh 30 5z Rez b 29 35 after introducing Dh = 4b for the channel.664. In this case it is approximately 2. given that the boundary layer develops more slowly in a channel flow.27. it will be shown that this approximate scaling result compares well with more exact solutions. f ReL = C2 L+ L .e. Care must be taken to ensure that fully developed flow prevails over most of a duct or channel.6 times greater. P. Thus. In the heat transfer and fluid flow literature the convention is to use the hydraulic diameter 4A / P or hydraulic radius 2A / P. L Lh Lh 27 Lh = C0. More recently. one can choose several combinations of these generic characteristics to obtain a length scale L= 4A . in many texts.asme. as the favorable pressure gradient in the inviscid core leads to much slower boundary layer development than the case for Blasius flow. respectively. the boundary layers on each plate in laminar flow will merge at a point z Lh downstream. NOVEMBER 2009 in addition to a host of other possibilities defined using the particular duct or channel dimensions.e. when L = Dh.e. then we shall refer to a short duct and long duct in the following manner: L L Lh Lh short duct long duct 28 Later.L. Thus it is reasonable to expect that the friction factor would be greater by approximately the same value. 15 .01 C0. i. This requires the use of exact values for accurate prediction of the pressure drop.org/terms/Terms_Use. One may easily get an estimate for the entrance length using simple boundary layer theory. 4. 27 when L+ = L+.328.L 34 This asymptotic behavior will be examined further and will form the basis for the new model developed for the hydrodynamic entrance problem.Dh 0. The actual solution for a channel is given by 4 Lh 0. If the entrance length is used as a criterion to measure the extent of the boundary layer region. whereas the traditional entrance length is defined on the basis of the centerline velocity. owing to the acceleration within the inviscid core. etc. PL.DhDh ReDh 32 where C0. its use in laminar flow has been questioned 7–9 .06. One issue that has not been adequately addressed in literature is the selection of an appropriate characteristic length scale. h C1 = or L+ = h C2 C1 2 C2 L+ h 33 C0. i. where the centerline velocity is 99% of the maximum velocity. Redistribution subject to ASME license or copyright. This characteristic length arises naturally from a simple control volume balance on an arbitrarily shaped straight duct. The choice of an appropriate characteristic length. It was determined that the widely used concept of the hydraulic diameter was inappropriate for laminar flow and the authors proposed using L = A as a characteristic length scale. 4. these are C2 = 1. a number of other Transactions of the ASME Downloaded 27 Oct 2009 to 134. A more detailed discussion and analysis on the use of L = A may be found in Refs. Given an arbitrary duct of crosssectional area A. 11. will now be examined. we easily see from Eqs. i. A simple equation relating the approximate magnitude of the hydrodynamic entrance length may be obtained using scaling analysis. which only marginally satisfies the entrance length criteria.4 Length Scales and Slenderness. 131. 13 . As will be seen shortly.3 Hydrodynamic Entrance Lengths. Merely stating that L Lh is not strictly valid for making an assumption of fully developed flow. L Lh. In an earlier work 10 . the boundary layer develops over a length four times greater than that predicted by the simple Blasius theory. The entrance length is traditionally defined as the point downstream of the inlet.. the entrance length may be written more generally as 111105-4 / Vol.44 for a mean friction factor and C2 = 1. experiences a significantly higher pressure drop than a fully developed flow. This measure of the development region is also utilized in the selection of an appropriate heat transfer model as well as ascertaining which type of friction model be utilized for pressure drop calculations.Dh varies weakly with duct shape. This solution represents the intersection of the asymptotic limits on a plot of the complete behavior of f Re versus L+. and 24 that the constant C2 1. 24 will be much larger than the value obtained for friction in a boundary layer flow over an isolated flat plate in streaming flow. The basis for the use of the hydraulic diameter arose in the modeling of turbulent flows in noncircular ducts. In turbulent flows. we have found from scaling analysis the following relationships for the friction factor–Reynolds number product: C1 . Depending on the channel or duct shape. Eq. once the exact limits are formulated. since a channel flow. For an isolated plate in streaming flow. when the hydraulic diameter is used. The constant C2 has been found theoretically by Siegel 6 for the circular duct to be C2 = 3. the velocity gradient varies only in a region very near the wall and there is deeper penetration of the flow field into re-entrant corners.

deficiencies in the hydraulic diameter concept should be addressed. But can be overcome if one defines 1 − Ai/Ao 1 + Ai/Ao 46 Clearly. a simple measure of slenderness is the ratio of the minor and major axes. In the case of the concentric annular duct. It is more fundamental than Eq. 42 . In the case of triangular ducts. while the radii ratio is a combination of parallel length scales. the characteristic length L = A Deff. and the duct length. a simple measure of its proportions is traditionally the ratio of inner and outer radii. In principle it exists as a convenient limit. it is never truly realized. If we consider the rectangular and elliptical ducts. The above formula is merely a generalized form where Ai / Ao = rà for the concentric annulus.27.org/terms/Terms_Use.153. the parallel plate is merely an idealization for low aspect ratio ducts and channels. the viscosity. 43 is more appropriate as it does not contain the transverse length scale. Other possibilities for L given above have a number of potential flaws. Finally. if one considers a duct of fixed area. NOVEMBER 2009. The issue becomes more complicated if one or both of the bounding ducts are polygons. A summary of the nominal aspect ratios of typical ducts is given in Table 1. This issue does not pose a problem for singly or doubly connected regions having finite area and perimeter.. as slenderness is a combination of transverse lengths scales. That is ¯ wA c = ¯ wA nc 39 Now if both the circular and noncircular ducts have the same mass ¯ flux G = w then Ac = Anc and the effective circular diameter becomes Deff = 4Anc 41 40 since it contains only the mass flow rate. Redistribution subject to ASME license or copyright. i. in practice. h b1 + b2 b1 + b2 . h 2h h/b h b/h h 1−r 1+r 1−r 1+r 1+e 1−r 1+r Aspect ratio 1 b a short long b1 + b2 /2 b1 + b2 /2 b b Trapezoid Isosceles triangle Annular sector ¯ w PL c 36 Circular annulus If the mean shear stress at the wall is constant. The use of perimeter may be considered from the point of view of constant shear stress or preservation of the retarding force and hence the surface area of the duct or channel. In all cases.e.. 131 / 111105-5 Downloaded 27 Oct 2009 to 134. This measure is not compatible with that of most other shapes we consider. However. However. since it has been established that L = A is more appropriate than the hydraulic diameter.cfm . better correlation of the laminar flow data was achieved versus an appropriate measure of duct slenderness. allowing a single approximate expression to be used for many duct shapes. 11 that L = A succeeds in bringing the dimensionless results closer together for similar ducts. but the duct aspect ratio → 0.e. Vol. That is ¯ w PL nc = Table 1 Definitions of aspect ratio Geometry Regular polygons Rectangle and ellipse Miscellaneous shapes 2h . the perimeter and area are not definable for the parallel plate channel. thus physically realizing the situation where A is a constant. the parallel plate channel is approached as the duct aspect ratio varied such that a b. short long 44 In other words. First. we should consider its effect on the dimensionless duct length L+. even this simple measure requires some additional attention for ducts with small re-entrant corners. it may also be argued on physical grounds that the square root of the flow area is essentially the same as preserving the cross-sectional duct area or maintaining a constant mass flux. However. i. It has been shown in Ref. the characteristic length L = P Deff.asme. rectangular and elliptical or polygonal 10–14 .160. Eq. then the following relationship holds: Pnc = Pc and the effective circular diameter becomes Deff = Pnc 38 37 Eccentric annulus In other words. that the hydraulically equivalent circular area and perimeter based on the hydraulic diameter are not the same as the true area and perimeter of the noncircular duct. see http://www. Journal of Fluids Engineering where Ai and Ao are the areas of the inner and outer polygons. namely. The measure of a duct’s slenderness can be easily defined for most simple shapes. This mismatch in area and perimeter is often cited as the probable cause in the mismatch of dimensionless laminar flow data 7 . Using Eq. 20 we may write either L+ = or L+ = L/ A L = Re A ˙ m 43 L/Dh LP2 = ˙ ReDh 16mA 42 such that a simple intrinsic value falling in the range 0 1 is obtained. This issue can be overcome by defining a simple measure of slenderness 1 − rà 1 + rà 45 This definition results from preserving the wall to wall spacing and defining a transverse length scale based on a rectangular duct of equivalent area. a measure of slenderness may defined as the ratio of base to height or height to base. Finally.

47 21.49 16.98 14.66 37.052 1.56 52.73 19. which provided 7 digit precision.004 1.1 Long Duct Asymptote. we may use the simpler expression.40 14.48 21.4 0.893 0. see http://www. considering only the first term of the series. and the circular annulus. Additionally.47 14. Also presented are values that result from using the asymptotic solution for the parallel plate channel. 4. where = b / a.76.90 1.76 18.83 21.07 17.038 1.49 15.98 14. Eq.20 0. the ratio of minor and major axes.17 .49 14.23 14.31 14.200 1.9 1 Exact 23. We now examine a number of important results employing both L = 4A / P and L = A.57 20.30 0.51 16.062 1. Starting with the elliptical duct. 11.49 15.06 17.59 20.84 14.38 14. 53 .16 19. Finally. The simplest duct shapes are the circular duct and the parallel plate channel.13.025 0.47 14.00 f Re f ReE Rectangular Elliptical 1.67 22. First. As seen in Table 3. This way.23 111.02 16.70 0.51 14.33 119.10 0. 24. 49 need not be evaluated.8 0.26 14. very little change is observed in the f Re values.75 16. the short duct limit is considered by re-examining the approximate solution obtained by Siegel 6 and Shapiro et al.074 1.21 17.13. Excellent correlation is achieved with the single term solution for the rectangular duct. The results for several other flat duct geometries are shown in Figs. 4 using a 30 term series. merely resulting from a Equation 51 has the following limits: f ReDh = 14.77 36. the long duct limit is considered and a simple expression developed for predicting the constant f Re= C1. 50 and 54 . The results for a wide range of aspect ratios are tabulated in Ref. Redistribution subject to ASME license or copyright.69 35. →1 →0 52 Table 3 †4‡ f Re results for elliptical and rectangular geometries f ReDh R f Re A Examination of the single term solution reveals that the greatest error occurs when = 1. These solutions have been cataloged in Refs.18 14. →1 →0 50 Next.028 1.28 14. If the solution is recast using L = A.60 19.37 15. asymptotic analysis 15 is used to establish expressions for the characteristic long duct and short duct behaviors established through scaling analysis.32 14. Thus.2 0.27.09 16.46 15.889 119. The solution for the dimensionless average wall shear 4 .40 14.909 0. Further. 51 and 53 for both characteristic length scales.17 19. the dimensionless average wall shear is found to be 4 f ReDh = 2 2 Equation 53 379.5 Asymptotic Analysis Table 2 Comparison of single term approximation for f Re f ReDh =b/a 0. in the limit of small aspect ratio.28 14.900 0.78 18. the results for the elliptic duct and the rectangular duct have also come closer together.60 0. the rectangular duct.82 16.6 0. and Equation 47 has the following limits: f ReDh = 16.35 49. 3 and 4.001 0. as a characteristic length scale.26 15.23 Equation 51 23.55 14. A sample is provided in Table 2 for comparison with the simpler one term approximation.20 14.38 14.18 f ReR f ReE 1.47 14.82 25.42 12. a comparison is made between the elliptical and rectangular duct solutions.00 53. which gives a f Re value 0.82 25.13 12/ 379.021 1.3 0.78 18. f Re A= →1 →0 54 1+ 2 2 12 E 47 = 1 − 2.60 17.00 14.77 36. Eqs.160.17 19.77 36. . NOVEMBER 2009 Transactions of the ASME Downloaded 27 Oct 2009 to 134. the maximum difference between the values for f Re occur in the limit of → 0.29 16.91 18.97 23.68 22.56 52.07 17.48 21. 49 and 53 shows that this difference is only 7. solutions for many ducts are examined.51 16.13 Exact 379.94 14.1 0.004 = 1+ 1− 12 192 5 53 tanh 2 Equation 53 now has the following limits: 111105-6 / Vol.01 24.147 1. L š Lh. 131.65 20.10 16.5 0.14 .org/terms/Terms_Use.005 1.81 25. Beyond this aspect ratio.48 16.55 14.34 13.153.68 22.7% below the exact value of f Re= 14. It is also clear from this analysis that the square root of cross-sectional area is more appropriate than the hydraulic diameter for nondimensionalizing the laminar flow data.50 0.010 1.34 15. 50 and 54 .23.97 15.7 0.31 18.01 0.00 23. →1 →0 48 If the solution is recast using L = A. 16 .55 14.007 1. Eqs. Next. 19. as a characteristic length scale. the following expression is obtained: f Re A = b / a Rectangular Elliptical 0.74 14.56 52.23 19. to compute values for the elliptical duct.890 0.096 1.47 14. the issue of an appropriate characteristic length scale is addressed and comparisons are made for both the hydraulic diameter Dh and square root of cross-sectional area A as characteristic length scales.23 f Re A In this section.7.47 120.12 16.cfm .05 0. In order to establish the long duct limit.924 0. Comparison of Eqs.61 14.18.65 12.947 0.7%.978 0. this asymptotic result does an adequate job of predicting the values of f Re A up to = 0.01 0.79 14.24 16. we examine the rectangular duct. It is now apparent that the solutions for the circular duct and square duct have essentially collapsed to a single value.61 14. the following relationship is obtained: f Re A = 2 3/2 1+ 2 E 49 Equation 49 now has the following limits: 8 f Re A= 2 3/2 14. Clearly.97 23.95 26. the elliptic integral in Eq.90 17.asme.97 16.59 20. Eqs. 5.05 0. These important shapes also appear as limits in the elliptical duct.84 14.014 1.37 15.33 119.80 0. gives f ReDh = 1+ 2 1− 24 192 5 51 tanh 2 Table 2 presents a comparison of the exact values 4 with the single term approximation.12 16.40 0.26 14.43 14.

Finally we examine some simple doubly connected regions.921 0.33 14. such that rà = ri / ro. 44 .004 0. When the characteristic length scale is changed to L = A.153.975 0.06 14.13 14.963 0. †4‡ Journal of Fluids Engineering NOVEMBER 2009. rà → 1 56 The annulus now has limits comparable to Eq.957 0. These are given in Table 1 and result from the intuitive definition of Eq. the appropriate aspect ratio as a measure of duct slenderness is also used. In other words. see http://www. we consider the regular polygons.org/terms/Terms_Use.941 0. The results are summarized in Table 4. 5 that excellent agreement is obtained when the results are rescaled according to L = A and a new aspect ratio defined as rà = Ai / Ao.33 f ReDh 16 for 3 N tween the triangular and the circular ducts is approximately 16.88 16 f Re A 15.993 1.60 15. Figure 5 shows data for polygonal annular ducts 4 along with the result for the circular annulus. data from Ref.05 15.asme.23 14. Values for f Re for the circular annulus and other shapes have been examined by Muzychka 11 . In addition to the change in characteristic length scale. †4‡ Fig. there is some departure from the circular annulus result Fig. 131 / 111105-7 Downloaded 27 Oct 2009 to 134. data from Ref.19 14.989 0. The solution for an annulus with inner radius ri and outer radius ri. It should be noted that as the inner boundary approaches the outer boundary.1% for the equilateral triangle.27.31 15. 45 is used as an effective aspect ratio. and less than 0.996 1.970 0.04 14. the following expression is obtained: f Re A change in the characteristic length scale.Table 4 f Re results for polygonal geometries †4‡ f ReP f ReC 0.833 0. the relative difference is reduced to 7. Redistribution subject to ASME license or copyright. which are essentially ducts of unity aspect ratio.990 0.990 0.23 14. As a result of defining an appropriate measure of slenderness and introducing the more appropriate characteristic length scale L = A. the data have collapsed onto a single curve. data from Ref.160.52 15. It is clear from Fig.05 14. which are bounded externally by a polygon or internally by a polygon 4 . 54 when Eq.889 0. there is little difference between a circular annulus and a rectangular duct. The results considered are those of the circular annulus and other annular ducts. Values for f ReDh fall in the range .071 1.03 14. à rà → 0 58 = 12 1+r .04 14.1% for the remaining polygons. The solution for a concentric annular duct is easily found in polar coordinates. rà → 0 24.41 15.991 0. and contains both the circular duct limit and the parallel plate channel limit.cfm . This definition was chosen since it returns the same rà ratio for the circular annular duct. Next. is found in most elementary fluids’ texts as follows: f ReDh = 16 1 − rà 2 1 + r − 1 − rÃ2 /ln 1/rà Ã2 =8 1 − rà 1 − rÃ2 1 + r − 1 − rÃ2 /ln 1/rà Ã2 57 Equation 57 has the following exact rà → 0 and approximate rà → 1 limits: 8 f Re A . The relative difference be13. rà → 1 1 − rà 55 Equation 55 has the following limits: f ReDh = 16.988 0. 4 f Re A for other noncircular ducts.7%.73 15. 5 f Re A for annular ducts. Vol. a result of their ability to be circumscribed by the circular duct. †4‡ If the solution is recast using the square root of the crosssectional area as a characteristic length scale.01 14.18 Fig.000 N 3 4 5 6 7 8 9 10 20 f ReDh 13. 3 f Re A for regular flat ducts.992 0.000 f ReP f ReC 1.

However.88 L+ L+ 1/2 1/2 tanh 2 69 + 111 L+ ¯ 63 Goldstein 19 obtained a solution for the core velocity using a method proposed in Schlichting 20 . If a local friction factor is desired.72. − 1 = 4f L D 6 59 Model Development and Comparisons Having developed the simple expressions for large and small values of the dimensionless duct length.16–8. Using the results provided by Eq.5 L+ + 1060 L+ 3/2 ¯ 62 = 3. Redistribution subject to ASME license or copyright. which further reduces the scatter of data points for ducts with re-entrant corners. Siegel 6 developed an expression relating the boundary layer displacement thickness to the velocity in the core. and U is the mean velocity. 5. i. It is now clear that Eq. lies in the range 1. the proposed model predicts all of the developing flow data available in literature to within 10% or better with few exceptions.. 53 fully characterizes the flow in the long duct limit for many ducts including doubly connected regions when the appropriate aspect ratio is used. A plot of over 700 data points in Ref. which we will denote as Lfd. 67 . 66 . in addition to models for the hydrodynamic entrance length L+. it is found that the value of n. the authors introduce the idea of an effective aspect ratio.44 L+ . uc is the velocity in the core. The solution begins with the definition of the short duct friction factor. Limiting values of rà are provided in Ref. The model takes the form yà = yà n + yà 0 f Re A n 1/n 1/n where u is the velocity distribution in the boundary layer. The most accurate method was the application of the method of Thwaites see Ref. the model when using the intrinsic aspect ratio yields very good agreement. see http://www.001. L+ → 0 L+ → 66a f Re A = 1+ 1− 12 192 5 . The maximum deviation of exact values is of the order 7–10%.88 L+ L+ 1/2 1/2 L D + 111 L+ L D 2 ¯ 64 where L+ is defined by Eq. which solves for the velocity in the core using a series expansion. a relationship for the dimensionless core velocity needs to be found. which are now functions of the duct geometry. The proposed model provides equal Transactions of the ASME Downloaded 27 Oct 2009 to 134.23 . Equation 65 is independent of the duct shape and may be used to compute the friction factor for the short duct asymptote of most noncircular ducts. Twentysix data sets were examined from Refs. 18 is similar to Fig. 2 .6 with a mean value n 2 11. 19 . The results of Goldstein 19 yield similar results. tanh 2 66b 4 4 uc + 0 D − 2y udy 60 we shall proceed to develop a simple model for the apparent friction factor as a function of L+. This requires simple rules for each shape. A similar analysis for the parallel channel 20 yields the same leading term as that for the circular duct.2 Short Duct Asymptote. Comparisons of the model are presented in Figs. L ™ Lh. These points have not been shown on the plot. and are summarized in Table 5.due to the flow field becoming multiply connected.44 is replaced with 1.17 . With the exception of the eccentric annular duct at large values of rà and eÃ.17 . Thus very near the inlet of any noncircular duct.160. which relates the friction factor to the velocity in the inviscid core pi − pL uc = 1 2 U U 2 2 f ReL = 3. the constant 3. which may be obtained by writing Bernoulli’s equation in the entrance region. Rearranging this expression leads to 4 U =1− uc D 1− 0 8 u dy + 2 uc D 1− 0 u ydy uc 67 68 61 Next.44 1 – 2.1 Developing Flow Friction Factor. 3.44 L + 2 + 1+ 1− 12 192 5 2 1/2 Substitution of the above result into the expression for the friction factor yields f ReD = 3.5 n 3. a crescent shape.asme.22. the effects of geometry become more pronounced and the solution for the circular duct is no longer valid. Siegel 6 then obtained the following four term approximation for the velocity in the core near the entrance of a circular duct: uc = 1 + 6. NOVEMBER 2009 Using the available data 4. As the boundary layer begins to grow further downstream. the following relationship may be written. Overall. An analytical result for the friction factor in the entrance region of the circular duct was obtained by Siegel 6 using several methods. Eq. but reduces the scatter to less than 3% in fully developed flows. the leading term of the solution is valid. In a recent paper by Duan and Yovanovich 18 . 6. The leading term in the solution for any characteristic length L is f ReL = 3. which minimizes the root mean square rms difference. with the leading term in the series being exactly the same.e. 131. This result represents a seminal achievement in nondimensionalization.44 L+ 65 which is valid for L+ = L / L ReL 0.27. using Pohlhausen’s approximate velocity distribution.e. and the general expression. with predictions lying in the range 1.44 1 – 2. f Re A In order to determine the friction factor.org/terms/Terms_Use. and a model for a new dimenh + sionless fully developed flow length.153.88 L+ U 1/2 = C2 L+ n + C1 n where n is a superposition parameter determined by comparison with numerical data over the full range of L+.. Analysis of the expressions developed by Siegel 6 and Goldstein 19 reveals that the leading term may be nondimensionalized using any characteristic length scale without introduction of scaling terms 111105-8 / Vol.cfm . Since the pressure gradient is only a function of axial position.72% rms error. 4. Siegel 6 applied several approximate analytical methods to obtain a solution for the velocity in the inviscid core. 6–8 for the most common duct shapes. i. rescaling the additional terms in the expression results in scaling parameters. 11 . the following model is proposed: f Re A − 43. The Siegel 6 analysis begins with the integrated form of the continuity equation in the entrance region where the boundary layer is small relative to the duct diameter D 2U = D−2 2 = 3. A general model is now proposed using the Churchill and Usagi 21 asymptotic correlation method.

the proposed model is now only a function of the dimensionless duct length L+ and aspect ratio . b / a = 1.30/4.63/5.44/7.77 5. For this geometry A → .53 3.6/5.3 1.9 10. 131 / 111105-9 Downloaded 27 Oct 2009 to 134.91 5.18 2.3 Eccentric annulus eà = 0.03 1.11/1.9.42/3. particularly for selecting an appropriate heat transfer model.1 Eccentric annulus eà = 0.73 1.69 6.4/1.14 2.03 1.9 0.5/5.14 1. data from Ref.60 1.9 2.16 1.62 1.89/8.3/2.08 8.2/0.50 1. the length Fig.71 2.27 1.1/0.08 7.35 2.12 1.63 5.7 1.38 3.72 5.35 rms 1.77 5.13/9.2 2.39 13. or better accuracy than the model of Yilmaz 2 and is also much simpler. A comparison of the model with the data for the parallel plate channel is also provided.1/6.14 2.89 1.5. 6 f Re A for developing laminar flow in polygonal ducts.16 2.9 1.04 3.2 Hydrodynamic Entrance Lengths.9 2.6/1.27 0.60/16.9. rà = 0.3/3.41 0.22/2.01 or a circular annular duct with rà 0.0/1.2 9.67 1.87 35.61/1.5 Ellipse b / a = 1 Ellipse b / a = 0. Vol.30 18.27 1.32 1.96 1.55 Optimal value.97 2.98/5.22 6.94 13.97/3.95 1.5 Ellipse b / a = 0.22/0.56 5.86 1.28/0.18 1.97 7. It is both simple and accurate for most engineering calculations for microchannel and minichannel systems.73/1. Finally.96/5.61 1.20 1.Table 5 Comparison of rms and percent − predicted… / „analytical… à 100… in developing flow models Shah 1978 min/max 1.51 1.7 17.77 1.14/12.09 1.48/2.61 1.70/5.17 1.76 1.56 3.66 7. or any other scale system where laminar flow prevails.59 9.06/1.68 2. However. this geometry may be accurately modeled as a rectangular duct with = 0. 7 f Re A for developing laminar flow in regular flat ducts.70 2.44 2.27.57 3.25 deg Circular sector 2 = 22.75 2.1/16 10.1 1.5 - differences „%diff= „analytical Proposed model Geometry Circle Circular annulus rà = 0. there is no need to introduce the term K .5/3. rà = 0.05 2.10 Circular annulus rà = 0.50 Circular annulus rà = 0.2 5. whereas the models of Shah 1 and Yilmaz 2 are functions of many more parameters.5.75 Square b / a = 1 Rectangle b / a = 0. it is desirable to develop a simple expression for the hydrodynamic entrance length. data from Ref.05 Circular annulus rà = 0. Shah 1 .04 1. 6.75 1.10 1.33 4.0 9. One notable feature of the new model is that it does not contain the incremental pressure drop term K .4/1.75 4.07 2.13/6.01 1. Good agreement is obtained with the current model when the parallel plate channel is modeled as a finite area duct with small aspect ratio.5 deg Circular sector 2 = 45 deg Circular sector 2 = 90 deg Pentagon Trapezoid = 72 deg.68 4.asme.73 1.79 6.75/7. Since the solution of Siegel 6 for the entrance region accounts for both the wall shear and the increase in momentum due to the accelerating core.123 a min/max 2.1 2. †4‡ Fig.5.cfm .5 Rectangle b / a = 0. Redistribution subject to ASME license or copyright.16 2.9 5. rà = 0.153.9/ 3.41 8.73 4. rà = 0.72 7.4/0.160.6 2.71 1.99 2.96 1.69 5. which appears in the models of Bender 3 .48/2.org/terms/Terms_Use.87 2.1 1. and Yilmaz 2 .4 2.2 17.4 Yilmaz 1990 min/max 0.98 2.34 2.2 Parallel plates b / a → 0 Isosceles triangle 2 = 30 deg Isosceles triangle 2 = 60 deg Isosceles triangle 2 = 90 deg Eccentric annulus eà = 0.10 0.66 1.7 2.0 1.04 1. Traditionally.46/12.97/1.5 Eccentric annulus eà = 0.97 1.75 2. †4‡ Journal of Fluids Engineering NOVEMBER 2009.62 3.8 1.3 2. see http://www. Thus. The hydrodynamic entrance length is useful for determining the extent to which the flow develops.69 2.01 2.72 3.7.01 3.2 Circular sector 2 = 11.1 Eccentric annulus eà = 0.85 2.0 1.29 10.4/1.15 2.6/1.86 2.41 1.05 n a n = 2 rms 1. rà = 0.96/11.71 1.1 11.85 7.

The present model predicts most of the developing flow data within 10% or better and 1.0211 0. The present study took advantage of scale analysis.0263 0. L+ = 0. along with associated conversions for the rectangular duct and elliptical duct when using the hydraulic diameter as a Table 6 Hydrodynamic entrance lengths Model L+ A = C0. asymptotic analysis.asme. The advanhy tage of Eq. h.1 0.0435 0. These lengths when rescaled to be based on the hydraulic diameter take the values L+ = 0. we obtain. They h compare well with the more exact results from Shah and London 4 for the circular duct. in such a manner that boundary layer region contributes little to the overall pressure drop.01.D h + Ellipse Lh. 53 be adopted. see http://www.00083 when = 0. which is related to the hydrodynamic entrance length. These expressions can be used to determine the size of the hydrodynamic entrance region for use with appropriate heat transfer models. Finally these models may also be used to predict results for ducts for which no solutions or tabulated data exist. A complete tabulation is provided in Table 6.0499 0. 66b for C1. This hydrodynamic length can be considered the required dimensionless duct length for which fully developed flow prevails. and for determining the extent of a duct for which entrance effects are negligible. and the selection of a more appropriate characteristic length scale to develop a simple model.5 0.16–8. 8 f Re A for developing laminar flow in the circular annular duct. Once again.00424 0. which state that fully developed flow pressure drop may be predicted using a Hagen–Poiseuille law result for f Re when L L h. NOVEMBER 2009 Downloaded 27 Oct 2009 to 134. A length scale.e. This model only requires two parameters: the aspect ratio of the duct and the dimensionless duct length. 70 . ˙ m A 71 Equation 74 clearly shows that a duct having a dimensionless duct length of order ten times the hydrodynamic entrance length is required for all boundary layer effects to be lost in the pressure drop calculation.0534 0. A 72 The above expression reduces to L+ = 0. Another useful expression may be developed from the asymptotic limits for determining the extent of a duct length.6 0. 53 with small error. Conveniently.8 0.0316 0.0582 0.160. In this case we propose a criterion that a duct length. and parallel plate channel. which can be applied to any duct shape.0459 A simple model was developed for predicting the friction factor–Reynolds number product in noncircular ducts for developing laminar flow.0387 0. Additional models were also developed using the asymptotic results for both the hydrodynamic entrance length and the fully developed flow length. when L = A.e. 72 is that it provides a single universal expression.0288 0.0586 0.D h 0. and also accounts for the aspect ratio effect. 131.01 0. f Re L+ This yields a result of + Lfd 1.0485 0. h L+ h or Lh = C0.153.021 when = 0.cfm . This is in marked contradiction to the many fluid mechanics texts.0564 0. 6–8 .0540 0.0449 0.0379 0. 33 with L+ = L+.72% rms for eight singly connected ducts and two doubly connected ducts.011. which is required for making a valid assumption of fully developed flow. and the model of Yilmaz 2 consists of several equations.059 when = 1 and L+ h h = 0.27.0404 0.0353 0. an approximate expression for the entrance length as a function of aspect ratio is obtained as follows: L+ = 0. This solution represents the intersection of the asymptotic limits on a plot of the complete behavior of f Re versus L+. whereas the model of Shah 1 requires tabulated values of three parameters.75 10L+ h 74 C2 C1 2 C0.0570 0.0326 0.01 for a channel. which tends to reduce entrance lengths for flatter ducts. one in which the pressure drop may be wholly determined using only Eq.0456 0.7 0.000828 0.0582 0. Transactions of the ASME 111105-10 / Vol. 0.99umax 4 .0181 0.org/terms/Terms_Use. This is often followed with an assumption of fully developed flow downstream of the hydrodynamic entrance length. Acknowledgment The authors acknowledge the financial support of the Natural Sciences and Engineering Research Council of Canada NSERC . in a fully developed laminar flow see Figs. 66a for C2 = 3.047 for a tube or square h duct when = 1 and L+ = 0.0434 0. †4‡ required for hydrodynamic boundary layer development in straight ducts of constant cross-sectional area is usually defined as the point where the centerline velocity is 0. which is 5% greater than that predicted by Eq.0269 0. It was also shown that this is possible. A 70 A Re A = C0. A new definition for the hydrodynamic entrance length may be obtained from Eq. Redistribution subject to ASME license or copyright. It should also be noted that a significant variation in data for L+ is found in literature.0453 0.0586 A + Rectangle Lh.05C1 C2 C1 2 73 9.0264 0.0822 1 + h 2 7 Summary and Conclusions 1− 192 5 2 tanh 2 = C0.0556 0.0440 0. it yields values that agree better than an order of magnitude guess. since the use of the of the square root of the cross-sectional flow area was a more effective characteristic length scale than the hydraulic diameter for collapsing the numerical results of geometries having similar shape and aspect ratio.05 0.00874 0. i.44 and Eq. data from Ref.. we see that the order of magnitude is h correctly and easily predicted.0361 0.2 0.056.Fig.0233 0.9 1 0.4 0. Substituting from Eq.. i.3 0. An equation relating the approximate magnitude of the hydrodynamic entrance length may be obtained by considering an equality between the two asymptotic limits given in Eq. h L+ = 0.0274 0. which yields a dimensionless pressure drop.

3. “Laminar Non-Newtonian Fluid Flow in Non-Circular Ducts and Micro-Channels. H. rad dynamic viscosity.. m/s average velocity. S. 2006. Vol. K.. Laminar Flow in Microchannel of Arbitrary Cross-Section. A.153. pp. m Superscripts C E P R ¯ · circle ellipse polygon rectangle mean value References 1 Shah. McGraw-Hill. UL / Reynolds number based on L velocity vector. H. 15 Leal. 23 Muzychka. Paper No. Laminar Flow and Convective Transport.. and Culham.cfm . “A Correlation for Laminar Hydrodynamic Entry Length Solutions for Circular and Non-Circular Ducts. and W.. pp. 2009. 1974. 682–686. “Slip Flow in Non-Circular MicroChannels. “General Equations for Pressure Drop for Laminar Flow in Ducts of Arbitrary Cross Sections. 177–179. National Congress of Applied Mechanics. b / a half angle... 15–19. Y. Aug. 19 R. 1990. and Kline. Wiley. 2002.. “Pressure Drop of Fully Developed.” ASME J. L / L ReL mass flow rate. MD. e / ro − ri 1 / 2 U2 Fanning friction factor 2 mass flux. pp. 1954. NM. N / m2 radius. ON. and Yovanovich. R. S.” Handbook of Single Phase Convective Heat Transfer. 2008. 1992. Modern Developments in Fluid Dynamics. Oxford. 128.D.160..” Second AIAA Theoretical Fluid Mechanics Meeting. 98-2492. A. m/s velocity components.. “Analytical and Experimental Study of Fluid Friction and Heat Transfer in Low Reynolds Number Flow Heat Exchangers. New York. Technol. 13 Duan. 1972. Edizioni ETS. M. thesis. m constant scaling constants linear dimension. Convection Heat Transfer. Energy Resour. T. m hydraulic diameter. 1978. Redistribution subject to ASME license or copyright. 1036–1044. “Modelling Friction Factors in Non-Circular Ducts for Developing Laminar Flow. P. New York. S. P. and Yovanovich.. Thonon.D. Siegel. eds.. “Friction Factor in the Laminar Entry Region of a Smooth Tube. 2 Yilmaz.. M. Clelata. “Solutions of Poisson Equation Within Singly and Doubly Connected Domains. 1979..” Ph. 17 Shah. C1 .. M. J.. Yovanovich. Shah. G. 14 Muzychka. Y. 10 Yovanovich. “Pressure Drop for Laminar Flow in Micro-Channels of Arbitrary Cross-Sections. M. 733–741.-Tech. m diameter of circular duct. 131 / 111105-11 Downloaded 27 Oct 2009 to 134. A. McGraw-Hill. K.z Greek Symbols boundary layer thickness. G. 1979.” ASME J. “Laminar Convective Heat Transfer in Ducts. Mar. “Druckverlust bei Laminarer Stromung im Rohreinlauf. New York. M. R. 18. pp. and Usagi. S. m/s Cartesian coordinates... CA.” Proceedings of the Second U. kg/s number of sides of a polygon correlation parameter perimeter. R. K. 15–18. m dimensionless duct length. Fluids Eng.. and Muzychka. University of Waterloo. Fluids Eng.. ed. L. 473–484. m nominal aspect ratio. Upper Saddle River. and Edge. 4 Shah. J. Boston. 130. kg/ m3 wall shear stress. S. 1121– 1128. M. Canada. 3 Bender. AIAA 97-3880.27. C2 c D Dh E · e eà f G K L L+ ˙ m N n P p r rà ReL ReL V u. MA. 1999. Y. MA. M. m dimensionless radius ratio.. 18 Duan.. Shah. and Muzychka. Italy. m pressure..” Sc. 12 Bahrami.y. 4A / P complete elliptic integral of the second kind eccentricity. 111201. Albuquerque. Kakac..asme. E. M. S. 7 White... 2nd ed. Nanofluid.. Kandlikar. McGraw-Hill.-Ing. New York.” Compact Heat Exchangers: A Festschrift on the 60th Birthday of Ramesh K. A. L. A.. S.. Cambridge. 1997. 1978.. “The Effect of Heating on Boundary Layer Transition for Liquid Flow in a Tube. pp. M.. ri / ro Reynolds number. Bontemps.w U x. M. 1987. 1995.” Microfluid. Goldstein. 5 Bejan. 3 4 . Fluid Mechanics. 1953. 112.b C C0 ... p.. and Bhatti.” 1997 National Heat Transfer Conference. pp. New York. S. Prentice-Hall. 123–130. pp. 9 Mironer. Process Fluid Mechanics. 41. W. M.S. B. 22 Muzychka. Oxford University Press. Massachusetts Institute of Technology. kg/ s m incremental pressure drop factor arbitrary dimension duct length. 11 Muzychka. eds. K.org/terms/Terms_Use. R. R. P. Wiley. 1980.. R. N / m2 Subscripts A c c Dh eff fd h i long L nc o short based on square root of flow area core circular based on hydraulic diameter effective fully developed hydrodynamic inner longest perpendicular based on the arbitrary length scale L noncircular outer shortest perpendicular fully developed limit flow area.. AIAA Paper No. 21 Churchill. F.” Chem. 8 Denn. Z.. Fluids Eng.. 100. Y. m dimensionless eccentricity. Jun.v. Laminar Flow Forced Convection in Ducts.” ASME J. and S. thesis. Y. 220–223.. 20 Schlichting. see http://www. M. S. 10–12. 1938. M..... 16 Shapiro. 6 Siegel. and Yovanovich. Academic. N s / m2 fluid density. pp. New York. M. Aung. S. 1969. R. NJ.. 1998.” IEEE Semi-Therm Symposium.Nomenclature A a. Chap. Baltimore.. R. J.” ASME J. San Jose. ButterworthHeinemann. Waterloo. 2007. “A General Expression for the Correlation of Rates of Transfer and Other Phenomena. Y... m2 major and minor axes of ellipse or rectangle. “Laminar Flow Friction and Heat Transfer in Non-Circular Ducts and Channels Part I Hydrodynamic Problem. and London. Viscous Fluid Flow. Z. Boundary Layer Theory.” AIChE J. Journal of Fluids Engineering NOVEMBER 2009.

Sign up to vote on this title
UsefulNot useful