1 Introduction
Turbulent boundary layers (TBLs) are among the most studied canonical fluid problems but most past studies are devoted to the flat plate (planar) TBL. A recent review by Smits, McKeon & Marusic (Reference Smits, McKeon and Marusic2011) describes the current understanding and future challenges of wall-bounded flows at high Reynolds number (
$Re$
). A variety of hydrodynamic engineering applications however, involve axisymmetric TBLs, which involve an additional length scale parameter to account for curvature. Several engineering applications have axisymmetric TBLs evolving under the influence of pressure gradients due to their geometrical shapes. For example, figure 1 shows a generic submarine hull (Groves, Huang & Chang Reference Groves, Huang and Chang1989) along with the streamwise varying pressure gradients experienced by the hull boundary layer.
Different pressure gradient conditions experienced by the streamwise (
$x$
) evolving hull boundary layer on a generic submarine hull, AFF8 (Groves et al.
Reference Groves, Huang and Chang1989).

The radius-based Reynolds number (
$Re_{a}=aU/\unicode[STIX]{x1D708}$
, where
$U$
is free-stream velocity,
$\unicode[STIX]{x1D708}$
is kinematic viscosity and
$a$
is the radius of cylinder) does not include any effect of wall shear stress or boundary layer thickness. Therefore, popular non-dimensional parameters to characterize axisymmetric TBLs are the ratio of boundary layer thickness to the radius of curvature (
$\unicode[STIX]{x1D6FF}/a$
) and the radius of curvature in wall units (
$a^{+}$
). Based on these two parameters, three regimes can be identified (Piquet & Patel Reference Piquet and Patel1999): (i) both
$\unicode[STIX]{x1D6FF}/a$
and
$a^{+}$
are large, (ii) large
$\unicode[STIX]{x1D6FF}/a$
and small
$a^{+}$
and (iii) small
$\unicode[STIX]{x1D6FF}/a$
and large
$a^{+}$
. The first flow regime is observed for axial flow over a long slender cylinder at high
$Re$
, where a large effect of curvature is felt. The second flow regime is realized for axial flow over slender cylinders at low
$Re$
, where axisymmetric TBLs behave like an axisymmetric wake with an inner layer with strong curvature and low
$Re$
effects. Almost all the experimental studies reported in the literature have focused on the first two regimes (see Piquet & Patel Reference Piquet and Patel1999). The third flow regime is common in applications where the Reynolds number is high but the boundary layer is thin compared to the radius of curvature. Usually, this flow regime is treated as a planar boundary layer where the curvature effects are assumed minimal. Although, there are significant fundamental differences between a planar TBL and a thin axisymmetric TBL at high
$Re$
, such as increased skin friction and rapid radial decay in turbulence away from the wall (Lueptow Reference Lueptow1990).
One of the earliest analytical investigation of the effect of transverse curvature on skin friction was conducted by Landweber (Reference Landweber1949), who used a
$1/7$
th power law for the velocity profile and the Blasius skin-friction law (Schlichting Reference Schlichting1968) to show that for a given momentum thickness (
$\unicode[STIX]{x1D703}$
) based Reynolds number (
$Re_{\unicode[STIX]{x1D703}}$
), axisymmetric boundary layers have higher skin friction and lower boundary layer thickness in comparison to planar boundary layers. Seban & Bond (Reference Seban and Bond1951) analysed the laminar boundary layer for axial flow over a circular cylinder from the governing boundary layer equations and showed that the skin-friction and heat-transfer coefficients for axisymmetric laminar boundary layers are higher than that obtained from the Blasius solution. Kelly (Reference Kelly1954) introduced an important correction to their solution, known as the Seban–Bond–Kelly (SBK) solution for zero pressure gradient (ZPG) axisymmetric boundary layers. The SBK solution was extended to the regime of large curvature effect as encountered in axial flow over long thin cylinders by Glauert & Lighthill (Reference Glauert and Lighthill1955). Stewartson (Reference Stewartson1955) provided an asymptotic solution for ZPG laminar axial flow over long thin cylinders.
Axisymmetric TBLs have not received the same attention as planar TBLs likely due to the inherent difficulties in keeping the flow perfectly axial and preventing sagging or elastic deformation of the cylinders. The effect of curvature has been the focus of most past studies. Richmond (Reference Richmond1957) and Yu (Reference Yu1958) conducted the first few experimental studies for curvature effects on boundary layers, which was followed by extensive experimental studies (Rao Reference Rao1967; Cebeci Reference Cebeci1970; Chase Reference Chase1972; Rao & Keshavan Reference Rao and Keshavan1972; Patel Reference Patel1974; Patel, Nakayama & Damian Reference Patel, Nakayama and Damian1974; Willmarth et al. Reference Willmarth, Winkel, Sharma and Bogar1976; Luxton, Bull & Rajagopalan Reference Luxton, Bull and Rajagopalan1984; Lueptow, Leehey & Stellinger Reference Lueptow, Leehey and Stellinger1985; Krane, Grega & Wei Reference Krane, Grega and Wei2010) showing that the transverse curvature indeed has a significant effect on the overall behaviour of axisymmetric TBLs.
Afzal & Narasimha (Reference Afzal and Narasimha1976) analysed thin axisymmetric TBLs at high
$Re$
(regime 3 described above) using asymptotic expansions and modified the well-known classical law of the wall for planar TBLs to include the effect of curvature. The wall-normal distance in wall units (
$y^{+}$
) was modified as,
where,
$a^{+}=au_{\unicode[STIX]{x1D70F}}/\unicode[STIX]{x1D708}$
is the radius of curvature in wall units. Using this modified
$y^{+}$
, it was shown that there exists a log layer in the mean velocity profile similar to that found in a planar TBL, with same slope but the intercept (
$B$
) is a weak function of curvature (
$B=5+236/a^{+}$
). It has been shown that
$U^{+}=a^{+}\ln (1+y/a)$
is valid in the viscous sublayer region, but the use of
$y^{+}$
from (1.1) instead of the planar
$y^{+}$
in the logarithmic region assumes that transverse curvature affects both the viscous sublayer and log layer identically.
One of the earliest numerical simulations of axisymmetric boundary layers were performed by Cebeci (Reference Cebeci1970), who showed higher skin friction compared to flat plate prediction in both laminar and turbulent regimes. Similar behaviour of skin friction was observed in numerous subsequent simulations of axisymmetric TBLs. Axisymmetric TBLs over long thin cylinders have been extensively studied by Tutty (Reference Tutty2008) using Reynolds-averaged Navier–Stokes (RANS) models and Jordan (Reference Jordan2011, Reference Jordan2013, Reference Jordan2014a ,Reference Jordan b ) using direct numerical simulations (DNS) and large eddy simulations (LES). Jordan used his simulation database to propose simple models for the skin friction (Jordan Reference Jordan2013) and the flow field (Jordan Reference Jordan2014b ).
None of the studies mentioned so far have considered pressure gradient effects. Experiments by Fernholz & Warnack (Reference Fernholz and Warnack1998) and Warnack & Fernholz (Reference Warnack and Fernholz1998) considered axisymmetric TBL under favourable pressure gradient (FPG) in internal flow.
Boundary layers under adverse pressure gradients (APG) have been studied in the past using asymptotic expansions (see Afzal (Reference Afzal1983, Reference Afzal2008) and references therein). Recently, Wei & Klewicki (Reference Wei and Klewicki2016) performed an integral analysis of the governing equations for ZPG boundary layers over flat plates and obtained,
where
$U_{e}$
and
$V_{e}$
are the mean streamwise and wall-normal velocities at the edge of the boundary layer, respectively,
$H$
is the shape factor and
$u_{\unicode[STIX]{x1D70F}}=\sqrt{\unicode[STIX]{x1D70F}_{w}/\unicode[STIX]{x1D70C}}$
is the friction velocity. The analysis was later extended for planar boundary layers under a pressure gradient by Wei, Maciel & Klewicki (Reference Wei, Maciel and Klewicki2017), which modified (1.2) as,
where
$\unicode[STIX]{x1D6FD}_{RC}$
is the Rotta–Clauser pressure gradient parameter (Rotta Reference Rotta1953; Clauser Reference Clauser1954),
$\unicode[STIX]{x1D6FF}^{\ast }$
is the displacement thickness and
$\unicode[STIX]{x1D6FF}$
is the boundary layer thickness.
$\unicode[STIX]{x1D6FD}_{RC}$
is often used to quantify the strength of the APG in boundary layer flows.
The goal of the present work is to analyse the governing equations of axisymmetric boundary layers evolving under the influence of a pressure gradient and understand the effect of transverse curvature on the flow. Integral analysis of the governing equations is performed in § 2 and the obtained relations are compared to the existing data in § 3. Implications of analytical relations are discussed in § 4. Section 5 concludes the paper.
2 Integral analysis of axisymmetric boundary layer
The boundary layer approximations for the time-averaged Navier–Stokes equations in cylindrical coordinates yield,
$$\begin{eqnarray}\displaystyle & \displaystyle rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}+rV\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}=-\frac{r}{\unicode[STIX]{x1D70C}}\frac{\text{d}P}{\text{d}x}+\frac{\displaystyle \unicode[STIX]{x2202}\left(r\unicode[STIX]{x1D708}\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}\right)}{\unicode[STIX]{x2202}r}+\frac{\unicode[STIX]{x2202}(-r\overline{u^{\prime }v^{\prime }})}{\unicode[STIX]{x2202}r}, & \displaystyle\end{eqnarray}$$
where
$U$
and
$V$
are mean, and
$u^{\prime }$
and
$v^{\prime }$
are fluctuations in axial and radial velocities respectively. Note that the stress term involving
$\unicode[STIX]{x2202}(\overline{u^{\prime }u^{\prime }}-\overline{v^{\prime }v^{\prime }})/\unicode[STIX]{x2202}x$
has been ignored on the right-hand side of (2.2) for the present analysis. This term however, cannot be neglected for large magnitude pressure gradients and boundary layers on the verge of separation. We have not made any assumption on the nature of boundary layer i.e. it can be laminar, transitional or turbulent. This implies that the present analysis holds as long as the governing equations (2.1), (2.2) are valid.
For a boundary layer under a pressure gradient, the mean wall-normal velocity outside the boundary layer (
$V_{o}$
) is not constant. Hence, the boundary layer equations are integrated in the wall-normal direction from the surface,
$r=a$
to a location outside the boundary layer,
$r=a+k\unicode[STIX]{x1D6FF}$
where
$a$
is the radius of curvature (cylinder),
$k\geqslant 1$
is a parameter and
$\unicode[STIX]{x1D6FF}$
is the boundary layer thickness. Note that setting
$k=1$
makes
$V_{o}=V_{e}$
, which is the mean wall-normal velocity at the edge of the boundary layer. Integration of (2.1) and (2.2) with the aforementioned limits yield,
$$\begin{eqnarray}\displaystyle \int _{a}^{a+k\unicode[STIX]{x1D6FF}}r\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r & = & \displaystyle -\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{\unicode[STIX]{x2202}(rV)}{\unicode[STIX]{x2202}r}\,\text{d}r=-(rV)|_{a}^{a+k\unicode[STIX]{x1D6FF}}\nonumber\\ \displaystyle & = & \displaystyle -(a+k\unicode[STIX]{x1D6FF})V_{o},\end{eqnarray}$$
$$\begin{eqnarray}\displaystyle \int _{a}^{a+k\unicode[STIX]{x1D6FF}}rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r+\int _{a}^{a+k\unicode[STIX]{x1D6FF}}rV\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}\,\text{d}r & = & \displaystyle -\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{r}{\unicode[STIX]{x1D70C}}\frac{\text{d}p}{\text{d}x}\,\text{d}r+\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{\displaystyle \unicode[STIX]{x2202}\left(r\unicode[STIX]{x1D708}\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}\right)}{\unicode[STIX]{x2202}r}\,\text{d}r\nonumber\\ \displaystyle & & \displaystyle +\,\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{\unicode[STIX]{x2202}(-r\overline{u^{\prime }v^{\prime }})}{\unicode[STIX]{x2202}r}\,\text{d}r\nonumber\\ \displaystyle & = & \displaystyle -\unicode[STIX]{x1D6FD}_{RC}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{2\unicode[STIX]{x1D6FF}^{\ast }}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}+\left.\left(r\unicode[STIX]{x1D708}\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}\right)\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-(r\overline{u^{\prime }v^{\prime }})|_{a}^{a+k\unicode[STIX]{x1D6FF}},\nonumber\\ \displaystyle & & \displaystyle\end{eqnarray}$$
where
$\unicode[STIX]{x1D6FD}_{RC}$
is defined as,
and
$f|_{a}^{b}=f(b)-f(a)$
. Using the boundary conditions,
the right-hand side of (2.4) can be evaluated. This yields,
$$\begin{eqnarray}\left.\begin{array}{@{}c@{}}\displaystyle \int _{a}^{a+k\unicode[STIX]{x1D6FF}}rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r+\int _{a}^{a+k\unicode[STIX]{x1D6FF}}rV\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}r}\,\text{d}r=-\unicode[STIX]{x1D6FD}_{RC}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{2\unicode[STIX]{x1D6FF}^{\ast }}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-au_{\unicode[STIX]{x1D70F}}^{2}\\ \displaystyle \;\Longrightarrow \;\int _{a}^{a+k\unicode[STIX]{x1D6FF}}rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r+(rVU)|_{a}^{a+k\unicode[STIX]{x1D6FF}}-\int _{a}^{a+k\unicode[STIX]{x1D6FF}}U\frac{\unicode[STIX]{x2202}(rV)}{\unicode[STIX]{x2202}r}\,\text{d}r=-\unicode[STIX]{x1D6FD}_{RC}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{2\unicode[STIX]{x1D6FF}^{\ast }}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-au_{\unicode[STIX]{x1D70F}}^{2}\\ \displaystyle \;\Longrightarrow \;\int _{a}^{a+k\unicode[STIX]{x1D6FF}}rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r+(a+k\unicode[STIX]{x1D6FF})V_{o}U_{e}+\int _{a}^{a+k\unicode[STIX]{x1D6FF}}rU\frac{\unicode[STIX]{x2202}U}{\unicode[STIX]{x2202}x}\,\text{d}r=-\unicode[STIX]{x1D6FD}_{RC}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{2\unicode[STIX]{x1D6FF}^{\ast }}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-au_{\unicode[STIX]{x1D70F}}^{2}\\ \displaystyle \;\Longrightarrow \;\int _{a}^{a+k\unicode[STIX]{x1D6FF}}r\frac{\unicode[STIX]{x2202}U^{2}}{\unicode[STIX]{x2202}x}\,\text{d}r=-(a+k\unicode[STIX]{x1D6FF})V_{o}U_{e}-\unicode[STIX]{x1D6FD}_{RC}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{2\unicode[STIX]{x1D6FF}^{\ast }}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-au_{\unicode[STIX]{x1D70F}}^{2}.\end{array}\right\}\end{eqnarray}$$
The shape factor,
$H$
is defined as,
Differentiating both sides with respect to
$x$
,
$$\begin{eqnarray}\displaystyle & \displaystyle \;\Longrightarrow \;H=\frac{\displaystyle \frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x}}{\displaystyle \frac{\text{d}\unicode[STIX]{x1D703}}{\text{d}x}}-\unicode[STIX]{x1D703}\frac{\displaystyle \frac{\text{d}H}{\text{d}x}}{\displaystyle \frac{\text{d}\unicode[STIX]{x1D703}}{\text{d}x}}. & \displaystyle\end{eqnarray}$$
Note that no assumption has been made regarding the self-similarity of the boundary layer as yet. The second term on the right-hand side of (2.14) is small as
$H$
varies very slowly with
$x$
as compared to
$\unicode[STIX]{x1D6FF}^{\ast }$
and hence, can be neglected. Self-similarity implies
$\text{d}H/\text{d}x=0$
, which makes the second term identically zero. Therefore,
$\unicode[STIX]{x1D6FF}^{\ast }$
and
$\unicode[STIX]{x1D703}$
for axisymmetric boundary layers are defined (Luxton et al.
Reference Luxton, Bull and Rajagopalan1984) such that,
Note that
$U=U_{e}$
for
$r\geqslant \unicode[STIX]{x1D6FF}$
, hence (2.16) and (2.17) can be written as,
since
$k\geqslant 1$
.
Differentiating both sides with respect to
$x$
and using the Leibniz integral rule on the right-hand side yield,
$$\begin{eqnarray}\displaystyle 2(\unicode[STIX]{x1D703}+a)\frac{\text{d}\unicode[STIX]{x1D703}}{\text{d}x} & = & \displaystyle \frac{2}{U_{e}}\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{\unicode[STIX]{x2202}(rU)}{\unicode[STIX]{x2202}x}\,\text{d}r-\frac{2}{U_{e}}\frac{\text{d}U_{e}}{\text{d}x}I\nonumber\\ \displaystyle & & \displaystyle -\,\frac{2}{U_{e}^{2}}\int _{a}^{a+k\unicode[STIX]{x1D6FF}}\frac{\unicode[STIX]{x2202}(rU^{2})}{\unicode[STIX]{x2202}x}\,\text{d}r+\frac{4}{U_{e}}\frac{\text{d}U_{e}}{\text{d}x}J,\end{eqnarray}$$
where,
Using (2.3) and (2.10) on the right-hand side of (2.20) and (2.21) yields,
$$\begin{eqnarray}\displaystyle & \displaystyle 2(\unicode[STIX]{x1D703}+a)\frac{\text{d}\unicode[STIX]{x1D703}}{\text{d}x}=-2\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})+2\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}I\hspace{108.00018pt} & \displaystyle \nonumber\\ \displaystyle & \displaystyle \hspace{60.00009pt}+\,2\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}+2a\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}-4\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}J & \displaystyle \nonumber\\ \displaystyle & \displaystyle \;\Longrightarrow \;2(\unicode[STIX]{x1D703}+a)\frac{\text{d}\unicode[STIX]{x1D703}}{\text{d}x}=2a\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}+2\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}I+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-4\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}J. & \displaystyle\end{eqnarray}$$
Dividing (2.24) by (2.25) and using (2.15) followed by rearranging of the terms, we get,
$$\begin{eqnarray}\displaystyle \left(\frac{\unicode[STIX]{x1D6FF}^{\ast }+a}{\unicode[STIX]{x1D703}+a}\right)H & = & \displaystyle \left[\frac{\displaystyle 2\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})-2\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}I}{\displaystyle 2a\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}+2\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}I+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\left.\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}r^{2}\right|_{a}^{a+k\unicode[STIX]{x1D6FF}}-4\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}J}\right].\quad\end{eqnarray}$$
Using the definitions of
$\unicode[STIX]{x1D6FF}^{\ast }$
(2.18) and
$\unicode[STIX]{x1D703}$
(2.19), it can be shown that,
Also, equation (2.24) yields,
$$\begin{eqnarray}\displaystyle & \displaystyle (\unicode[STIX]{x1D6FF}^{\ast }+a)\frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x}=\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})-\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}I & \displaystyle \nonumber\\ \displaystyle & \displaystyle \;\Longrightarrow \;\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})=(\unicode[STIX]{x1D6FF}^{\ast }+a)\frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x}+\frac{\unicode[STIX]{x1D6FD}_{RC}}{2\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}(r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}-r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}). & \displaystyle\end{eqnarray}$$
Hence, equation (2.26) can be rearranged to show that,
$$\begin{eqnarray}\displaystyle & & \displaystyle \left(\frac{\unicode[STIX]{x1D6FF}^{\ast }+a}{\unicode[STIX]{x1D703}+a}\right)H\left[2a\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}(r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}-r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }})+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}\right.\nonumber\\ \displaystyle & & \displaystyle \left.\qquad -\,2\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}(r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}-r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}-r^{2}|_{a}^{a+\unicode[STIX]{x1D703}})\right]\nonumber\\ \displaystyle & & \displaystyle \quad =2\frac{V_{o}}{U_{e}}(a+k\unicode[STIX]{x1D6FF})-\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}(r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}-r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }})\end{eqnarray}$$
$$\begin{eqnarray}\displaystyle & & \displaystyle \;\Longrightarrow \;2\frac{V_{o}U_{e}}{u_{\unicode[STIX]{x1D70F}}^{2}}(a+k\unicode[STIX]{x1D6FF})\left(\frac{\unicode[STIX]{x1D703}+a}{\unicode[STIX]{x1D6FF}^{\ast }+a}\right)=H\left[2a+\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}(r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}+2r^{2}|_{a}^{a+\unicode[STIX]{x1D703}})\right]\hspace{40.00006pt}\nonumber\\ \displaystyle & & \displaystyle \qquad \hspace{40.00006pt}+\,\left(\frac{\unicode[STIX]{x1D703}+a}{\unicode[STIX]{x1D6FF}^{\ast }+a}\right)\frac{\unicode[STIX]{x1D6FD}_{RC}}{\unicode[STIX]{x1D6FF}^{\ast }}(r^{2}|_{a}^{a+k\unicode[STIX]{x1D6FF}}-r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}).\end{eqnarray}$$
Substituting for
$V_{o}$
from (2.29) and rearranging,
$$\begin{eqnarray}\displaystyle (\unicode[STIX]{x1D703}+a)\frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x} & = & \displaystyle H\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}}\left[a+\frac{\unicode[STIX]{x1D6FD}_{RC}}{2\unicode[STIX]{x1D6FF}^{\ast }}(r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}+2r^{2}|_{a}^{a+\unicode[STIX]{x1D703}})\right]\nonumber\\ \displaystyle \;\Longrightarrow \;\frac{u_{\unicode[STIX]{x1D70F}}^{2}}{U_{e}^{2}} & = & \displaystyle \frac{\displaystyle (\unicode[STIX]{x1D703}+a)\frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x}}{\displaystyle H\left[a+\frac{\unicode[STIX]{x1D6FD}_{RC}}{2\unicode[STIX]{x1D6FF}^{\ast }}(r^{2}|_{a}^{a+\unicode[STIX]{x1D6FF}^{\ast }}+2r^{2}|_{a}^{a+\unicode[STIX]{x1D703}})\right]}\nonumber\\ \displaystyle \;\Longrightarrow \;C_{f} & = & \displaystyle \frac{\displaystyle 2\left(1+\frac{\unicode[STIX]{x1D703}}{a}\right)\frac{\text{d}\unicode[STIX]{x1D6FF}^{\ast }}{\text{d}x}}{\displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right]}.\end{eqnarray}$$
Self-similarity of boundary layers implies that
$\unicode[STIX]{x1D6FF}^{\ast }/\unicode[STIX]{x1D6FF}$
is constant. So
$C_{f}$
can be written as,
$$\begin{eqnarray}\displaystyle C_{f}=\frac{\displaystyle 2\left(1+\frac{\unicode[STIX]{x1D703}}{a}\right)\frac{\unicode[STIX]{x1D6FF}^{\ast }}{\unicode[STIX]{x1D6FF}}\frac{\text{d}\unicode[STIX]{x1D6FF}}{\text{d}x}}{\displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right]}. & & \displaystyle\end{eqnarray}$$
Note that
$C_{f}=2u_{\unicode[STIX]{x1D70F}}^{2}/U_{e}^{2}$
is related to
$\unicode[STIX]{x1D6FD}_{RC}$
by definition (see (2.5)). But that definition contains external flow parameters. On the other hand, equation (2.33) relates
$C_{f}$
to the boundary layer parameters directly. Also, equation (2.31) can be rearranged to show that,
$$\begin{eqnarray}\displaystyle \frac{U_{e}V_{o}}{u_{\unicode[STIX]{x1D70F}}^{2}}\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)\left(1+k\frac{\unicode[STIX]{x1D6FF}}{a}\right) & = & \displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right.\nonumber\\ \displaystyle & & \displaystyle \left.+\,\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)\left(k\frac{\unicode[STIX]{x1D6FF}}{\unicode[STIX]{x1D6FF}^{\ast }}-1+\frac{k^{2}\unicode[STIX]{x1D6FF}^{2}-\unicode[STIX]{x1D6FF}^{\ast 2}}{2a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right].\qquad\end{eqnarray}$$
At the edge of the boundary layer,
$k=1$
and
$V_{o}=V_{e}$
. Therefore,
$$\begin{eqnarray}\displaystyle \frac{U_{e}V_{e}}{u_{\unicode[STIX]{x1D70F}}^{2}}\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)\left(1+\frac{\unicode[STIX]{x1D6FF}}{a}\right) & = & \displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right.\nonumber\\ \displaystyle & & \displaystyle \left.+\,\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)\left(\frac{\unicode[STIX]{x1D6FF}}{\unicode[STIX]{x1D6FF}^{\ast }}-1+\frac{\unicode[STIX]{x1D6FF}^{2}-\unicode[STIX]{x1D6FF}^{\ast 2}}{2a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right].\qquad\end{eqnarray}$$
At the verge of separation,
$u_{\unicode[STIX]{x1D70F}}$
goes to zero. Using the definition of
$\unicode[STIX]{x1D6FD}_{RC}$
(2.5), equation (2.35) yields,
$$\begin{eqnarray}\displaystyle V_{e} & = & \displaystyle -\unicode[STIX]{x1D6FF}^{\ast }\frac{\text{d}U_{e}}{\text{d}x}\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)^{-1}\left(1+\frac{\unicode[STIX]{x1D6FF}}{a}\right)^{-1}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right.\nonumber\\ \displaystyle & & \displaystyle \left.+\,\left(\frac{1+\unicode[STIX]{x1D703}/a}{1+\unicode[STIX]{x1D6FF}^{\ast }/a}\right)\left(\frac{\unicode[STIX]{x1D6FF}}{\unicode[STIX]{x1D6FF}^{\ast }}-1+\frac{\unicode[STIX]{x1D6FF}^{2}-\unicode[STIX]{x1D6FF}^{\ast 2}}{2a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right].\end{eqnarray}$$
3 Comparison to previous work
3.1 Consistency with planar boundary layer relations
For a planar boundary layer,
$1/a$
approaches
$0$
as
$a$
approaches
$\infty$
. Setting
$1/a=0$
in (2.33) and (2.34) yields,
$$\begin{eqnarray}\displaystyle C_{f}=\frac{\displaystyle 2\frac{\unicode[STIX]{x1D6FF}^{\ast }}{\unicode[STIX]{x1D6FF}}\frac{\text{d}\unicode[STIX]{x1D6FF}}{\text{d}x}}{H+\unicode[STIX]{x1D6FD}_{RC}(2+H)},\end{eqnarray}$$
and
At the verge of separation,
$u_{\unicode[STIX]{x1D70F}}=0$
; setting
$k=1$
yields,
These relations are identical to those derived by Wei et al. (Reference Wei, Maciel and Klewicki2017) (equations (13) and (14) of their paper) for a planar boundary layer with a pressure gradient. They compared their analytical relations to the data available in the literature for APG TBLs and found good agreement (see figures 2–5 of their paper).
Setting
$\unicode[STIX]{x1D6FD}_{RC}=0$
in (3.2) yields,
Note that for
$\unicode[STIX]{x1D6FD}_{RC}=0$
, regardless of the value of
$k$
,
$V_{o}$
is the same i.e.
$V_{o}=V_{e}$
is constant outside the boundary layer. Equation (3.4) was derived by Wei & Klewicki (Reference Wei and Klewicki2016) (equation (11) of their paper) and shown to be valid for laminar, transitional and turbulent boundary layers.
3.2 Axisymmetric ZPG laminar boundary layer
The SBK solution (Seban & Bond Reference Seban and Bond1951; Kelly Reference Kelly1954) for axisymmetric laminar boundary layer is valid up to
$\unicode[STIX]{x1D708}x/Ua^{2}<0.04$
, and was subsequently extended by Glauert & Lighthill (Reference Glauert and Lighthill1955) (GL) to the interval
$0.04<\unicode[STIX]{x1D708}x/Ua^{2}<100$
. For ZPG laminar axisymmetric boundary layer, equation (2.33) becomes,
$\unicode[STIX]{x1D6FF}^{\ast }$
can be obtained from either SBK or GL solutions and
$H=2.59$
for a laminar boundary layer. Thus,
$C_{f}$
can be obtained. Figure 2(a) shows
$C_{f}$
as a function of
$\unicode[STIX]{x1D708}x/Ua^{2}$
for three different
$Re_{a}=10\,000$
, 1000 and 500, compared with both SBK and GL solutions. Note that the difference in
$C_{f}$
using
$\unicode[STIX]{x1D6FF}^{\ast }$
from either solution (SBK or GL) is negligible. Our results smoothly transitions from the SBK to the GL solution as
$\unicode[STIX]{x1D708}x/Ua^{2}$
increases, as evident in the lower
$Re_{a}$
cases. Figure 2(b) compares our result with the numerical solution of Cebeci (Reference Cebeci1970), where
$Re_{a}$
is varied.
$\unicode[STIX]{x1D6FF}^{\ast }$
and
$H$
for this case are estimated from the asymptotic results of Stewartson (Reference Stewartson1955). The
$C_{f}$
obtained from the Blasius solution (
$C_{f}\sqrt{Re_{x}}=0.664$
) (Schlichting Reference Schlichting1968) is also shown for comparison. Overall, our results show good agreement with Cebeci (Reference Cebeci1970) for the entire range from thin to thick axisymmetric laminar boundary layers. Note that at large
$Re_{a}$
,
$\unicode[STIX]{x1D6FF}/a$
approaches zero and hence, the axisymmetric laminar boundary layer approaches planar behaviour.
Skin-friction coefficient (
$C_{f}$
) as a function of non-dimensional parameter
$\unicode[STIX]{x1D708}x/Ua^{2}$
(a), where results for radius based Reynolds number
$Re_{a}=500$
, 1000 and 10 000 are shown along with the solutions of Seban–Bond–Kelly (Seban & Bond Reference Seban and Bond1951; Kelly Reference Kelly1954) (▪ (red)) and Glauert–Lighthill (Glauert & Lighthill Reference Glauert and Lighthill1955) (▫). The present result (3.5) using
$\unicode[STIX]{x1D6FF}^{\ast }$
from SBK (——(red)) and GL (- - -), show identical
$C_{f}$
.
$C_{f}$
as a function of
$Re_{a}$
is compared with the result of Cebeci (Reference Cebeci1970) (▫) for a long thin cylinder (large
$x/a$
), where the boundary layer thickness reaches an asymptotic value (Stewartson Reference Stewartson1955) (b). The value obtained from the Blasius solution (- - -) is also shown in (b) for comparison.

3.3 Axisymmetric ZPG turbulent boundary layer
Cebeci (Reference Cebeci1970) numerically solved incompressible turbulent ZPG axial flow over a circular slender cylinder of radius,
$a=1^{\prime \prime }$
and
$Re_{a}=40\,200$
. The same relation (3.5) is used to estimate
$C_{f}$
but the
$C_{f,planar}$
correlation of Monkewitz, Chauhan & Nagib (Reference Monkewitz, Chauhan and Nagib2008) is used. The shape factor
$H$
is assumed to be 1.4 and the boundary layer growth
$\text{d}\unicode[STIX]{x1D703}/\text{d}x$
is assumed identical in both planar and axisymmetric cases. Figure 3(a) shows our results compared to that of Cebeci (Reference Cebeci1970). Note that the range of
$Re_{\unicode[STIX]{x1D703}}$
on the cylinder is large (
$1000<Re_{\unicode[STIX]{x1D703}}<10\,000$
). Hence, the assumption of identical growth and
$H=1.4$
may not hold, which is the reason for the difference between our result and that of Cebeci (Reference Cebeci1970). In reality,
$H$
is a weakly decreasing function of
$Re_{\unicode[STIX]{x1D703}}$
for TBLs (Monkewitz et al.
Reference Monkewitz, Chauhan and Nagib2008). For example,
$H\approx 1.45$
at
$Re_{\unicode[STIX]{x1D703}}=1000$
(Schlatter & Örlü Reference Schlatter and Örlü2010), whereas
$H\approx 1.36$
at
$Re_{\unicode[STIX]{x1D703}}=9000$
(Österlund Reference Österlund1999). The results shown in figure 3(a) will further improve if the variation of
$H$
with
$Re_{\unicode[STIX]{x1D703}}$
is taken into account.
Kumar & Mahesh (Reference Kumar and Mahesh2016) simulated thin axisymmetric TBLs in the range
$1400<Re_{\unicode[STIX]{x1D703}}<1620$
. Using their boundary layer
$\unicode[STIX]{x1D6FF}^{\ast }$
and
$\unicode[STIX]{x1D703}$
variation with streamwise distance
$x$
, which is almost linear, their slope
$\text{d}\unicode[STIX]{x1D6FF}^{\ast }/\text{d}x$
and
$\text{d}\unicode[STIX]{x1D703}/\text{d}x$
can be estimated. This estimated slope can be used to compute
$C_{f}$
for
$1500<Re_{\unicode[STIX]{x1D703}}<3000$
, as shown in figure 3(b). Our results are compared with the correlation of Monte, Sagaut & Gomez (Reference Monte, Sagaut and Gomez2011), which corrected the correlation of Woods (Reference Woods2006) using their extensive simulation database, showing good agreement. Note that for a large range of
$Re_{\unicode[STIX]{x1D703}}$
, the assumption of linear growth of boundary layer breaks down, hence the differences at large
$Re_{\unicode[STIX]{x1D703}}$
.
(a) Skin-friction coefficient (
$C_{f}$
) as a function of
$Re_{\unicode[STIX]{x1D703}}$
is compared with the result of Cebeci (Reference Cebeci1970) (▫) for a slender cylinder for radius based Reynolds number
$Re_{a}=40\,200$
and radius
$a=1^{\prime \prime }$
. The shape factor of
$H=1.4$
and
$C_{f,planar}$
correlation of Monkewitz et al. (Reference Monkewitz, Chauhan and Nagib2008) is used in our relation to predict
$C_{f}$
. The boundary layer growth is assumed identical to that of a flat plate, which need not be true for slender cylinders at high
$Re_{\unicode[STIX]{x1D703}}$
. (b)
$U_{e}^{+}$
as a function of
$Re_{\unicode[STIX]{x1D703}}$
is compared with the correlations of Monte et al. (Reference Monte, Sagaut and Gomez2011) and Woods (Reference Woods2006).
$U_{e}^{+}$
is related to
$C_{f}$
as
$U_{e}^{+}=\sqrt{2/C_{f}}$
.

4 Discussion
4.1 Effect of curvature on
$C_{f}$
If both planar and axisymmetric boundary layers have the same boundary layer parameters, equations (2.33) and (3.1) yield:
$$\begin{eqnarray}\displaystyle & \displaystyle \frac{C_{f,axisymmetric}}{C_{f,planar}}=\frac{\displaystyle \left(1+\frac{\unicode[STIX]{x1D703}}{a}\right)[H+\unicode[STIX]{x1D6FD}_{RC}(2+H)]}{\displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right]} & \displaystyle \nonumber\\ \displaystyle & \displaystyle \;\Longrightarrow \;\frac{C_{f,axisymmetric}}{C_{f,planar}}-1=\frac{\displaystyle \frac{\unicode[STIX]{x1D703}}{a}H+\unicode[STIX]{x1D6FD}_{RC}\left[\frac{\unicode[STIX]{x1D703}}{a}+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}\left(1-\frac{H}{2}\right)\right]}{\displaystyle H+\unicode[STIX]{x1D6FD}_{RC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right]}. & \displaystyle\end{eqnarray}$$
Thus, if the right-hand side of (4.1) is positive, the presence of curvature increases
$C_{f}$
and vice versa.
It is easy to see that for ZPG (
$\unicode[STIX]{x1D6FD}_{RC}=0$
) boundary layers,
For boundary layer with APG (
$\unicode[STIX]{x1D6FD}_{RC}>0$
), the denominator of the right-hand side of (4.1) is always positive. Hence, the effect of curvature will depend on the sign of the numerator
$\unicode[STIX]{x1D702}$
defined as,
It can be shown that
$\unicode[STIX]{x1D702}\geqslant 0$
if
$\unicode[STIX]{x1D6FD}_{RC}\geqslant 0$
(see appendix A). Therefore, the presence of curvature increases
$C_{f}$
if
$\unicode[STIX]{x1D6FD}_{RC}\geqslant 0$
. Note that this is true regardless of the value of
$a$
. It has been assumed that
$\text{d}\unicode[STIX]{x1D6FF}/\text{d}x$
is identical for both planar and axisymmetric TBLs. This is not always true. In fact, for a thick axisymmetric TBL at zero pressure gradient (
$\unicode[STIX]{x1D6FF}/a\gg 1$
and
$\unicode[STIX]{x1D6FD}_{RC}=0$
),
$\text{d}\unicode[STIX]{x1D6FF}/\text{d}x$
is smaller than that of the planar TBL value (Tutty Reference Tutty2008). However,
$C_{f}$
is still higher than planar values because
$\unicode[STIX]{x1D703}/a\gg 1$
, which compensates for the decrease in
$\text{d}\unicode[STIX]{x1D6FF}/\text{d}x$
.
The presence of curvature may or may not increase
$C_{f}$
in FPG axisymmetric TBLs depending on the sign of the right-hand side of (4.1).
4.2 Thick axisymmetric ZPG turbulent boundary layer
For
$\unicode[STIX]{x1D6FD}_{RC}=0$
, the expression for
$C_{f}$
(2.33) reduces to,
Thus, knowing local boundary layer parameters,
$C_{f}$
can be estimated. For example, Jordan (Reference Jordan2014a
) compiled numerous experimental results along with his simulation database for thick axisymmetric TBLs in ZPG and showed that
$\unicode[STIX]{x1D6FF}/\unicode[STIX]{x1D703}\approx 7.2$
. The estimated value is
$\text{d}\unicode[STIX]{x1D6FF}/\text{d}x\approx 2.5\times 10^{-3}$
for a range of thick axisymmetric TBLs (
$2.1\leqslant \unicode[STIX]{x1D6FF}/a\leqslant 11$
,
$37\leqslant a^{+}\leqslant 388$
,
$586\leqslant Re_{a}\leqslant 7475$
). This makes,
4.3 Axisymmetric TBL under large APG
For large APG,
$\unicode[STIX]{x1D6FD}_{RC}\gg 1$
. Thus (2.33) yields,
$$\begin{eqnarray}C_{f}\approx \left[\frac{\displaystyle 2\left(1+\frac{\unicode[STIX]{x1D703}}{a}\right)\frac{\unicode[STIX]{x1D6FF}^{\ast }}{\unicode[STIX]{x1D6FF}}\frac{\text{d}\unicode[STIX]{x1D6FF}}{\text{d}x}}{\displaystyle 2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{2a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)}\right]\frac{1}{\unicode[STIX]{x1D6FD}_{RC}}.\end{eqnarray}$$
For a self-similar TBL in an APG,
$\unicode[STIX]{x1D6FF}^{\ast }/\unicode[STIX]{x1D6FF}$
,
$H$
and
$\text{d}\unicode[STIX]{x1D6FF}/\text{d}x$
become constant (Maciel, Rossignol & Lemay Reference Maciel, Rossignol and Lemay2006). Similar behaviour is expected for an axisymmetric TBL as well. When
$\unicode[STIX]{x1D6FF}/a<1$
,
$\unicode[STIX]{x1D703}/a$
and
$\unicode[STIX]{x1D6FF}^{\ast }/a$
are small as compared to 1. This makes, the term inside brackets ([ ]) nearly constant. Thus for a thin axisymmetric TBL at large APG,
$C_{f}\sim 1/\unicode[STIX]{x1D6FD}_{RC}$
. A similar result was obtained by Wei et al. (Reference Wei, Maciel and Klewicki2017) for a planar TBL.
4.4 Axisymmetric TBL under FPG
For FPG TBLs, there are two important flow parameters: pressure gradient parameter (
$\unicode[STIX]{x1D6EC}$
) (Narasimha & Sreenivasan Reference Narasimha and Sreenivasan1973) and acceleration parameter (
$K$
) (Launder Reference Launder1964) defined as,
All the relations derived in § 2 hold for a FPG axisymmetric TBL as well by replacing
$\unicode[STIX]{x1D6FD}_{RC}$
with
$-\unicode[STIX]{x1D6EC}$
. It can be shown that,
$$\begin{eqnarray}\displaystyle \frac{\text{d}C_{f}}{\text{d}\unicode[STIX]{x1D6EC}}=C_{f}\left[\frac{\displaystyle 2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)}{\displaystyle H-\unicode[STIX]{x1D6EC}\left[2+H\left(1+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}+\frac{\unicode[STIX]{x1D703}^{2}}{a\unicode[STIX]{x1D6FF}^{\ast }}\right)\right]}\right]<0. & & \displaystyle\end{eqnarray}$$
Thus, increasing the FPG decreases
$C_{f}$
and this effect is expected to be enhanced by the presence of transverse curvature as the presence of terms with
$1/a$
enhances the magnitude of
$\text{d}C_{f}/\text{d}\unicode[STIX]{x1D6EC}$
.
5 Conclusion
In this work, an integral analysis of equations governing axisymmetric boundary layer flow is presented, including the effect of pressure gradient. Analytical relations are derived relating
$C_{f}$
to the boundary layer parameters. The relations for planar TBLs with and without a pressure gradient presented by Wei et al. (Reference Wei, Maciel and Klewicki2017) and Wei & Klewicki (Reference Wei and Klewicki2016) respectively can be recovered by setting
$1/a=0$
and further setting
$\unicode[STIX]{x1D6FD}_{RC}=0$
. It has been shown that the presence of transverse curvature increases
$C_{f}$
regardless of the nature of the boundary layer, consistent with the observations reported in the literature for both ZPG and APG axisymmetric boundary layers. The derived relations are compared to the existing results in the literature, showing good agreement. The results presented in this work are expected to be valid for any boundary layer as long as the governing equations hold, which assumes local dynamic equilibrium. It is challenging, both experimentally and computationally, to obtain accurate
$C_{f}$
at high
$Re$
. However, it is relatively easier to obtain accurate mean velocity profiles. In addition to predicting the influence of the pressure gradient and curvature, the derived expressions are potentially useful to both skin-friction measurements and wall-modelled large eddy simulation of turbulent boundary layers.
Acknowledgements
This work is supported by the United States Office of Naval Research (ONR) under ONR Grant N00014-14-1-0289 with Dr K.-H. Kim as technical monitor. We thank Mr S. Anantharamu for useful discussions.
Appendix A. Maximum value of
$\unicode[STIX]{x1D702}$
It is known that
$H\geqslant 1$
which yields,
Adding (A 1) and (A 2) we get,
$$\begin{eqnarray}\displaystyle \frac{H}{2}-1-\frac{1}{H}\geqslant -\frac{3}{2},\;\Longrightarrow \;\frac{1}{\displaystyle \frac{H}{2}-1-\frac{1}{H}}\leqslant -\frac{2}{3}. & & \displaystyle\end{eqnarray}$$
But,
$$\begin{eqnarray}\displaystyle \frac{1}{\displaystyle \frac{H}{2}-1-\frac{1}{H}}=\frac{H}{\displaystyle H\left(\frac{H}{2}-1\right)-1}=\frac{\displaystyle -\frac{\unicode[STIX]{x1D703}}{a}H}{\displaystyle \frac{\unicode[STIX]{x1D703}}{a}+H\frac{\unicode[STIX]{x1D703}}{a}\left(1-\frac{H}{2}\right)}=\frac{\displaystyle -\frac{\unicode[STIX]{x1D703}}{a}H}{\displaystyle \frac{\unicode[STIX]{x1D703}}{a}+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}\left(1-\frac{H}{2}\right)}. & & \displaystyle\end{eqnarray}$$
From (A 3) and (A 4), it follows that,
$$\begin{eqnarray}\displaystyle \frac{\displaystyle -\frac{\unicode[STIX]{x1D703}}{a}H}{\displaystyle \frac{\unicode[STIX]{x1D703}}{a}+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}\left(1-\frac{H}{2}\right)}\leqslant -\frac{2}{3}. & & \displaystyle\end{eqnarray}$$
Now,
$$\begin{eqnarray}\displaystyle & \displaystyle \unicode[STIX]{x1D702}=\frac{\unicode[STIX]{x1D703}}{a}H+\unicode[STIX]{x1D6FD}_{RC}\left[\frac{\unicode[STIX]{x1D703}}{a}+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}\left(1-\frac{H}{2}\right)\right]>0 & \displaystyle \nonumber\\ \displaystyle & \displaystyle \;\Longleftrightarrow \;\unicode[STIX]{x1D6FD}_{RC}>\frac{\displaystyle -\frac{\unicode[STIX]{x1D703}}{a}H}{\displaystyle \frac{\unicode[STIX]{x1D703}}{a}+\frac{\unicode[STIX]{x1D6FF}^{\ast }}{a}\left(1-\frac{H}{2}\right)}. & \displaystyle\end{eqnarray}$$
Using (A 5), it is easy to see that (A 6) always holds for
$\unicode[STIX]{x1D6FD}_{RC}>0$
.















