Transition Velocity Estimation for Flow in Large Pipes
P.T. Slatter Cape Peninsula University of Technology, South Africa
ABSTRACT
There is considerable pressure to operate slurry pipeline systems at higher solids concentrations, causing
the viscous character of the resulting slurry to become increasingly non-Newtonian and paste like in nature.
Prediction of the transition velocity is fundamentally important for design and operation. The objective of
this paper is to extend previous work to the more general Herschel-Bulkley rheological model. Using five
predictive approaches in literature and dimensionless groups, it is shown that the problem is particularly
acute for large pipes, and is yield stress dominated.
1 INTRODUCTION
There is considerable pressure from economic and environmental quarters to run slurry lines at higher solids
concentrations, resulting in paste flow. The viscous behaviour of these fine particle slurries becomes
increasingly non-Newtonian and paste like in nature at high concentration. Paste slurries have distinct
laminar and turbulent flow regimes, and prediction of the transition velocity is fundamentally important for
design and operation. Flow behaviour, pressure loss, wear rates and flow patterns are distinctly different in
the two regimes. Previously, the Bingham Plastic model was used to describe the paste rheology, as this is
the simplest model which exhibits a yield stress (Slatter and Wasp, 2000; Slatter, 2005). The objective of this
paper is to extend this approach to the more general Herschel-Bulkley rheological model. The presence of a
yield stress causes plug flow in the centre of a pipe. Plug flow is essentially solid type behaviour and as such
has a profound effect on the laminar/turbulent transition (Slatter, 1999). Dimensionless groups for
investigating this type of flow behaviour will be introduced and used for developing a definition for the term
“large pipe”. Various approaches for the estimation of the laminar/turbulent transition are reviewed and
compared in the context of these dimensionless groups, for large pipes.
2 LITERATURE AND THEORY
2.1 The Herschel-Bulkley rheological model
The Herschel-Bulkley rheological model (Govier and Aziz, 1972) can be formulated in terms of shear stress
τ;
τ τ y K γn (1)
or viscosity η;
τ τy
η Kγn 1 (2)
γ γ
The two terms on the right-hand side of Equation (2) will be equal when the boundary shear rate γb reaches a
value of:
1
τy n
γb (3)
K
where:
y = yield stress (Pa).
K = fluid consistency index (Pa.sn).
n = flow behaviour index.
γ = shear rate (s-1).
The importance of the boundary shear rate b is that it marks the boundary between yield stress and power
law domination of viscosity. This is shown graphically in Figure 1.
100
10
y Yield Stress
Dominated
Power Law
Dominated
b
Viscosity
Slope = n-1
0.1
0.01 Slope = -1
0.001
11 10 100 1000 10000 100000 1000000
Shear Rate
Figure 1 Graphical presentation of Equation (2) showing the boundary shear rate b and the
oblique asymptotes
It is significant to note from Figure 1 that the ordinate intercept of the oblique asymptote is in fact the yield
stress value τy. This illustrates the fact that – for practical purposes – the viscosity values for shear rates less
than the boundary shear rate b are directly proportional to the yield stress. This serves to emphasise the
importance of the yield stress value when operating in the region to the left of the boundary shear rate. The
yield stress τy is a strong function of particle properties, concentration and solution chemistry, with the
plastic viscosity K usually being a weaker function of these.
The Oldroyd number (No) is a dimensionless group, useful for characterising the pipe flow of these materials,
and is formulated as follows:
τy
NO
R n
(4)
K V
2.2 The Laminar/Turbulent Transition
Traditionally, the issue of the prediction of the critical velocity at which pipe flow changes from laminar to
turbulent, is dealt with using a Reynolds number. This paradigm has been carried through to non-Newtonian
pipe flow. We shall consider five approaches:- the wall viscosity Newtonian Reynolds number, the Metzner-
Reed Reynolds number, the Re3 Reynolds number approach developed by the author, the transition criterion
approach of Ryan and Johnson, and the Intersection Method.
The wall viscosity Newtonian Reynolds number uses the standard Newtonian Reynolds number with wall
viscosity as the denominator:
ρVD
Re Newt = (5)
μ0
Metzner and Reed (1955) developed a generalised Reynolds number for non-Newtonian flow:
8ρ V 2
Re MR n'
(6)
8V
K'
D
The computation of the Metzner and Reed rheological constants is somewhat tedious, and can be avoided by
calculating their Reynolds number directly from the laminar flow friction factor:
16
Re MR (7)
f lam
Slatter (1999) developed a Reynolds number that places emphasis on the yield stress. Using the fundamental
definition that Re α inertial / viscous forces, the final formulation is:
2
8ρ Vann
Re 3 n
(8)
8V
τ y K ann
D shear
where Vann and Dshear apply only to the sheared portion of the flow (ibid).
For the above three Reynolds number approaches, the transition criterion has been taken as Re=2100.
Ryan and Johnson (1959) and Hanks (1981) have derived stability functions for laminar flow velocity vector
fields. For axially symmetrical pipe flow the two functions differ by a factor of two. The Ryan and Johnson
stability function is:
R u ρ du
Z
τ 0 dr
(9)
The maximum value of this function Zmax across a given laminar velocity vector field is taken as the stability
criterion. For Newtonian flow, Zmax = 808 corresponds to Re = 2100 and it is assumed that all fluids will
obtain this value of Zmax = 808 at the transition limit. Although only the Ryan and Johnson criterion is
considered here, it must be noted that the analyses and general findings and conclusions applicable to the
Ryan and Johnson criterion apply equally to the Hanks approach. Clearly, the stability criterion approach is
the most scientifically sophisticated of all the approaches considered.
The intersection method proposed by Hedström is a practical approach which uses the intersection of the
laminar and turbulent flow theoretical lines as the critical point (Hedström, 1952; Shook and Roco, 1991;
Wilson and Thomas, 2006). The success of this method depends on the accuracy of the turbulent model used.
Hedström proposed that the Newtonian model be used. This model tends to under-predict and the Wilson and
Thomas model has been used here, as this model has given good results as reported by Xu et al. (1993).
3 PROBLEM ANALYSIS
The five approaches to be considered can be compared graphically, over a range of flow behaviour index (n)
values, using the Metzner-Reed Reynolds number as the reference standard. These are presented in
Figures 2-4, together with experimental data from Appendix A.
10000
Re3
n = 0.94
Metzner-Reed Reynolds Number
1000
ReNewt
Int.Meth
Zmax
100
0.1 1 10 100
Oldroyd Number
Figure 2 Model comparison and data for n=0.94
10000
Int.Meth Re3
Metzner-Reed Reynolds Number
1000
ReNewt
Zmax
n = 0.62
100
0.1 1 10 100
Oldroyd Number
Figure 3 Model comparison and data for n = 0.62
10000
Int.Meth Re3
Metzner-Reed Reynolds Number
1000
ReNewt
Zmax
n = 0.36
100
0.1 1 10 100
Oldroyd Number
Figure 4 Model comparison and data for n = 0.36
4 DISCUSSION
Conceptually, the Oldroyd number represents the ratio of the two terms on the RHS of Equation (1) in the
Herschel-Bulkley rheological model, with the value of (V/R) as a representative shear rate. At an Oldroyd
number of unity, the two terms are equal. For NO <1, the power law term dominates, and for NO>1, the yield
stress dominates. Figures 2 – 4 show that for NO <1, there is little difference in the approaches, whereas for
NO>1, in the region where the yield stress dominates, the different approaches diverge sharply.
Experimental data from Appendix A are shown on Figures 2 – 4, which show that the data agree best with
the Re3 Reynolds number approach. The significant difference between the data and the other approaches is
possibly due to the failure of these approaches to adequately accommodate the effect of the yield stress.
It is clear from Figures 2 – 4 that an Oldroyd number of unity is fundamentally and practically important. For
NO <1, there is little difference in the approaches, indicating that any prediction method could be used with
confidence. For NO>1, the problem becomes yield stress dominated, and it appears that only the Re 3
Reynolds number approach can be used – but more data is needed to verify this. In a conceptual sense, an
Oldroyd number of unity could be used as a border between “small” and “large” pipe behaviour.
It is believed that the data shown in Figures 2 – 4 are typical of the highest Oldroyd number data available.
Given the industrial context of the problem, it is important to extend the range of experimental values to
include practical site values i.e. for NO>100. This represents an important experimental and conceptual
frontier, which is required to be pushed back. From a practical design perspective, Figures 2 – 4 indicate that
at NO>100, order of magnitude errors can be expected when predicting critical velocities using the various
methods presented here – a fact upon which the literature is currently largely silent. It is hoped that this
present work will make a contribution in this direction.
5 CONCLUSIONS
It has been argued that the laminar/turbulent transition is of great importance for the pipe system designer,
and presents a particularly acute problem for large diameter pipes conveying yield stress fluids. Using the
Herschel-Bulkley rheological model as the point of departure, five of the approaches in literature have been
brought to bear on the problem. It has been shown that at low Oldroyd number, these approaches agree
closely, while at high Oldroyd number, these approaches diverge sharply. Only the Re3 approach agrees well
with the experimental data presented, at the higher Oldroyd numbers. Further research focussed on higher
Oldroyd numbers is called for.
REFERENCES
Govier, G.W. and Aziz, K. (1972) The Flow of Complex Mixtures in Pipes, van Nostrand Reinhold Co.
Hanks, R.W. (1981) Course notes. Hydraulic design for flow of complex fluids, 1981 Richard W Hanks Associates,
Inc., Orem, Utah. USA.
Hedström, B.O.A. (1952) Flow of plastics materials in pipes, Industrial and engineering chemistry, Vol. 44, No.3.
Metzner, A.B. and Reed, J.C. (1955) Flow of non-Newtonian fluids - correlation of the laminar, transition and turbulent
flow regions, AIChE Journal, Vol.1, No.4.
Ryan, N.W. and Johnson, M.M. (1959) Transition from laminar to turbulent flow in pipes, AIChE Journal, Vol.5
pp. 433-435.
Shook, C.A. and Roco, M.C. (1991) Slurry flow: principles and practice, Butterworth-Heinemann.
Slatter, P.T. (1999) The role of rheology in the pipelining of mineral slurries, Min. Pro. Ext. Met. Rev., Vol. 20,
pp. 281-300.
Slatter, P.T. (2005) Tailings Transport – Back to Basics! Invited Keynote Address, Paste 2005, International Seminar on
Paste and Thickened Tailings, Santiago, Chile 20-22 April, 2005. ISBN 0-9756756-3-X, pp. 165-176.
Slatter, P.T. and Wasp, E.J. (2000) The laminar/turbulent transition in large pipes. 10th International Conference on
Transport and Sedimentation of Solid Particles - Wrocław: 4-7 September 2000, pp. 389-399. ISBN 83-87866-
12-1.
Wilson, K.C. and Thomas, A.D. (2006) Analytical model of laminar-turbulent transition for Bingham plastics. Can. J.
Chem. Eng. 84, pp. 520–526.
Xu, J., Gillies, R., Small, M. and Shook, C.A. (1993) Laminar and turbulent flow of kaolin slurries, 12th Int. Conf. on
slurry handling and pipeline transport, Hydrotransport 12, BHR Group, p. 595.
APPENDIX A – CRITICAL VELOCITY DATA
Experimental data from the Flow Process Research Centre at the Cape Peninsula University of Technology,
South Africa.
Pipe Slurry Critical Slurry Rheological
Test ID
Diameter Density Velocity Parameters
D ρ Vc y K n
mm kg/m3 m/s Pa Pa.sn
K1.BRS 57.7 1284 7.7 119 0.045 0.94
K1.BRM 81.2 1284 7.7 119 0.045 0.94
K1.BRL 151 1284 7.4 119 0.045 0.94
K2.BRS 57.7 1316 8.4 153 0.791 0.62
K2.BRM 81.2 1316 8.2 153 0.791 0.62
K3.BRS 57.7 1351 8.6 201 5.910 0.36
