Flow Equations
The Pipe Flow Interface calculates the pressure and velocity of an incompressible or weakly compressible fluid by solving the continuity and momentum equations outlined below.
A one dimensional pipe can be present on a boundary in a 2D geometry, or on an edge in a 3D geometry.
Momentum and Continuity Equations
The momentum and continuity equations for flow in a pipe are given by (Ref. 16):
(2-1)
and
(2-2)
The second term on the right-hand side in Equation 2-1 represents the pressure drop due to viscous shear. Here, u is the cross-section averaged velocity (SI unit: m/s), ρ the density (SI unit: kg/m3), p pressure (SI unit: Pa), fD (dimensionless) the Darcy friction factor (see Expressions for the Darcy Friction Factor) and F is a volume force term (SI unit: N/m3).
Furthermore, dh is the mean hydraulic diameter (SI unit: m), given by:
(2-3)
where A is the pipe cross section area (SI unit: m2) available for flow, and Z is the wetted perimeter (SI unit: m).
Figure 2-5: Unit tangent vector to the pipe axis.
Let
be the unit tangent vector to the pipe axis.
If the following assumptions apply:
the momentum balance in Equation 2-1 can be rewritten as
(2-4)
and we can define the tangential velocity u as u = uet. The pipe flow physics interface in COMSOL Multiphysics solves for the tangential velocity u. This is also the quantity the user specifies in for example inflow boundary conditions and in the initial conditions settings.
Curvature pressure drop effects can be added in points with the Bend feature.
Gravity can be included explicitly in the model, but when the variation in density is negligible, and the model is not pressure driven (the inlet boundary condition is a pressure setting), the only effect of including gravity is a change in the total pressure level. In those cases, it is therefore common modeling practice to set the gravity F to 0 and interpret the pressure variable as the reduced pressure p = p − ρg(z0 − z), where z0 is the datum level of the free liquid surface. This reduces the model complexity and yields the same results.
The Darcy friction factor in Equation 2-1 accounts for the continuous pressure drop along a pipe segment due to viscous shear, and is expressed as a function of the Reynolds number (Re) and the surface roughness divided by the hydraulic diameter (e/dh).
(2-5)
where
(2-6)
The physics interface automatically calculates fD from one of the predefined expressions Equation 2-8 through Equation 2-14.
Additional Flow Resistances
Additional flow resistance due to irreversible turbulent effects in joints and junctions can be added as point conditions between pipe segments — Bends, Valves, T-junctions, Y-junctions, n-way junctions, Contractions, and Expansions. These resistances give rise to abrupt (lumped) energy losses due to turbulence in the points where they are added. The lumped pressure drop is calculated as:
(2-7)
For more information about each type, see the interface descriptions in The Pipe Flow Interface section.
Loss coefficients, Ki, for turbulent flow are available in the literature (Ref. 14) and the set predefined in the Pipe Flow interface is reproduced in the table below:
0.1 and 2.1 for side and main. See Figure 2-1 for a description of flow branches in a T-junction.
Above, β is the ratio of small to large cross-sectional area. The point friction losses listed above apply for Newtonian fluids. Point losses applying to non-Newtonian flow can be added as user-defined expressions, for instance from (Ref. 15).
For the point pressure loss features Bends, Valves, T-junctions, Y-junctions, n-way junctions, Contractions, and Expansions, the pressure loss can also be specified as a pressure drop Δp directly.
Expressions for the Darcy Friction Factor
Newtonian Fluids
For single-phase fluids, the Churchill equation (Ref. 1) for the Darcy friction factor can be used for the full range of Re (laminar, transition and turbulent) and full range of e/d:
(2-8)
where
(2-9)
(2-10)
In the laminar regime (Re < 2000), fD is independent of the surface roughness and is given by the Stokes formula:
(2-11)
The Wood equation (Ref. 2) gives the friction factor for 4000 < Re <  1·107 and 1·10-5 < e/d <  0.04, according to
(2-12)
with
(2-13)
The Haaland equation (Ref. 3) for the Darcy friction factor is commonly used for oil pipelines and wells. It can recover both small and large relative roughness limits for a wide range of Reynolds numbers (4·103 < Re < 1·108)
(2-14)
It can be rewritten as
(2-15)
For very low relative roughness e/d, the Haaland equation simplifies to Colebrook’s explicit formula (Ref. 4)
(2-16)
For very large relative roughness e/d, the Haaland equation simplifies to the von Karman formula (Ref. 5)
(2-17)
An alternative to Haaland equation is the Swamee-Jain equation (Ref. 6)
(2-18)
The equation is valid for relative roughness 1·10-6 < e/d <  1·10-2 and for Reynolds number in the range 5·103 < Re <  1·108.
All the above equations are selectable from a list of friction factor expressions. As noted, only the Churchill equation covers both the laminar and turbulent flows, as well as the transitional region in between these flow regimes. The equations by Wood, Haaland, Colebrook, von Karman, or Swamee-Jain, intended for the turbulent regime, are combined with the Stokes equation for laminar flow to cover all flow conditions. When Re<1000, COMSOL Multiphysics selects Stokes equation if it predicts a friction factor greater than equations for the turbulent regimes. This produces more accurate results that using the turbulent frictions factors in the laminar regime, but it does not necessarily produce accurate estimates of the friction factor in the transition region. Therefore, it is always advisable to check the Reynolds number for a give pipe flow solution and change the friction model and recalculate, if necessary.
The Newtonian fluid type also has two Gas-Liquid options, which employs a simple two-phase approach presented by Balasubramaniam and others (Ref. 25). This is a one-fluid approach that treats the fluid as one phase (one mass and momentum balance is solved), but it corrects the pressure drop correlation with empirical factors for liquid-gas mixtures. As such, the model also makes the following assumptions:
The first selectable method is the gas-liquid, friction factor multiplier, which modifies the single-phase Newtonian Darcy friction factor defined in Equation 2-8 and on, such that
(2-19)
where fD,L is the Newtonian single phase friction factor computed as described earlier in this section. fD is then used as usual in the pressure drop correlation in Equation 2-4. The two-phase friction factor is calculated as (Ref. 25)
(2-20)
where ρL, μL, ρG, and μG are the density and viscosity of the liquid, and gas, respectively, and where ωG is the quality (the gas phase mass fraction). The total density in Equation 2-20 is calculated as
(2-21)
or
(2-22)
where is the user provided volumetric void fraction. The user has a choice to either enter gas phase mass fraction ωG or volumetric void fraction . The conversion formula between the two is
(2-23)
The second selectable gas-liquid method is the gas-liquid, effective Reynolds number. This method uses an effective adjusted viscosity to calculate the Reynolds number in the pressure loss calculations. The default mixture viscosity model is the extended Einstein, which probably is the most accurate within a large range of gas mass fractions (Ref. 26):
(2-24)
and the other options are
Linear average
(2-25)
Reciprocal average
(2-26)
Liquid viscosity
(2-27)
User defined.
Non-Newtonian Power-law Fluids
For Power-law fluids the apparent viscosity is related to the shear rate as
(2-28)
where m and n are two empirical curve fitting parameters known as the fluid consistency coefficient and the flow behavior index, respectively. In the laminar regime the friction factor for power law fluids can be calculated by the Stokes equation using the modified Reynolds number proposed by Metzner and Reed (Ref. 7):
(2-29)
(2-30)
For the turbulent regime Irvine (Ref. 8) proposed the following expression for the friction factor:
(2-31)
where
(2-32)
Ryan and Johnson (Ref. 9) formulated a criterion for the transition between laminar and turbulent flow, where
(2-33)
predicts the critical Reynolds number.
Non-Newtonian Bingham Plastic Model.
The Bingham model describes viscoplastic fluid with a yield stress:
(2-34)
The yield stress, τB (SI unit: Pa), and the plastic viscosity, μB (SI unit: Pa·s), are found by curve fitting to experimental data.
The Swamee-Aggarwal equation (Ref. 10) gives the friction factor for a Bingham plastic fluid in the laminar regime according to:
(2-35)
where
(2-36)
and
(2-37)
is the Hedström number.
For turbulent flow Darby (Ref. 11) provided the equation:
(2-38)
with
(2-39)
as well as an equation covering all flow regimes:
(2-40)
where
(2-41)
The non-Newtonian friction models outlined above apply for pipes with circular cross-section. The generalized Reynolds number needs to be modified to account for other cross-sections.
Surface Roughness
Values of the absolute surface roughness found in the literature (Ref. 13 and Ref. 14) are reproduced in the table below: