Proceedings of OMAE 01

th
20 International Conference on Offshore Mechanics and Arctic Engineering
June 3-8, 2001, Rio de Janeiro, Brazil

OFT 01-1364
FLUID DYNAMIC LOADING ON CURVED RISER PIPES
Anthi Miliou, Spencer J. Sherwin, J. Michael R. Graham
Department of Aeronautics, Imperial College of Science, Technology and Medicine
London SW7 2BY, U.K.
e-mail: a.miliou@ic.ac.uk, s.sherwin@ic.ac.uk, m.graham@ic.ac.uk

ABSTRACT
In order to gain a preliminary understanding of the fluid
dynamics developed past a curved riser pipe, a numerical
investigation into the flow past curved cylinders at a Reynolds
number of 100 has been performed. To approximate the flow
conditions on curved riser pipes, different velocity profiles and
flow directions were applied and the corresponding results
compared. In addition, the fluid dynamic loading and the wake
structures for curved cylinder flows were investigated.
The fully three-dimensional simulations were computed
with a spectral/hp element method. The computational results
were compared with experiments undertaken in the towing tank
facility of the Department of Aeronautics of Imperial College.
INTRODUCTION
As the need to develop deep water reservoirs for the
exploitation of hydrocarbons below the sea has been increasing,
riser pipes are now required to operate in depths up to 2000m.
Catenary risers, which are effectively more flexible riser pipes,
are preferred over vertical risers for deep water reservoirs as
they can accommodate larger motions of the floating vessel.
While steel catenary risers are normally associated with large
radii of curvature, flexible catenary risers can have quite drastic
localised deformation. A catenary riser is subject to fluid
dynamic loading as a combination of waves and current and
may be in tangential contact with the seabed. In this case part of
its length lies near horizontal at the seabed being semi- or even
fully- buried occasionally.
Understanding the unsteady flow patterns developed
behind a deformed catenary riser pipe is a problem that requires
investigation in three dimensions. The empirical models
(Morison's equation, Shear 7 [5]) applied on a two-dimensional
sectional basis in commercial use, fail to predict accurately the
behaviour of a catenary riser at the touch down point, where the
riser is in contact with the seabed. This is a very sensitive
region for the structure particularly when the riser is being

lifted off and dropped back onto the seabed as a result of vessel
motion. The need to develop the computational means for
modelling and understanding the fluid dynamic loading,
especially in the lower region in proximity to the ground, is
therefore very important. The computational investigations
described in the current work have focused on gaining a better
understanding of the dynamics of the flow past curved
cylindrical pipes and their wake structures.
COMPUTATIONAL METHOD
Computations have been carried out using Nektar [1, 8,
10], a three-dimensional incompressible Navier-Stokes solver,
based on the spectral/hp element method. This spectral method
is a high-order discretisation method where one can increase
simultaneously the number of elements (h-refinement) and the
order, P, of the interpolating polynomials in the domain (prefinement). This method is associated with the method of
weighted residuals with the test functions being the same as the
basis functions. Unstructured conforming discretisations have
been used in the computational domain. This spectral/hp
element method accounts for exponential convergence.
The mesh generation for all the computational
investigations was accomplished with Felisa where a modified
advancing layers method [3] is employed near the pipe wall
regions and a method based on the advancing front technique
[4] is employed for the rest of the domain. Hybrid meshes were
constructed where the boundary layer mesh was composed of
prismatic elements growing in the direction of normal to the
pipe surface and the interior of the domain was filled with
tetrahedra.
RESULTS AND DISCUSSION
The first set of computations has been performed at a
Reynolds number, based on the diameter, of 100 and is aimed
at investigating the effects of curvature compared to a straight
cylinder and the effects of the interaction with the seabed as

Copyright #### by ASME

well as a sheared current profile. The computations included

cases of a quarter turn with either a horizontal or a vertical
extension, subject to either uniform or sheared flow in
directions normal or parallel to the plane of curvature.
Symmetry boundary conditions were assigned to the plane
boundaries at the intersection with the curved cylinder.
Specifically, the five cases presented in this paper include:
a quarter turn in uniform flow normal to the plane of
curvature
a deformed quarter turn with a vertical addition of two
diameters being semi-buried in the ground and subjected to
shear normal flow
a deformed quarter turn with a vertical addition of two
diameters being semi-buried in an inviscid wall and subjected
to shear normal flow
a deformed quarter turn with a vertical addition of two
diameters being semi-buried in an inviscid wall and subjected
to uniform normal flow
a quarter turn with a horizontal extension of ten diameters in
uniform flow parallel to the plane of curvature
The selection of the above cases was done on the basis that
a generic shape for a catenary riser would consist effectively of
a quarter turn with the vertical extension to the offshore vessel
and the horizontal extension lying on the seabed partially- or
fully- buried in the ground.
This work represents a fundamental study of the influence
of the seabed and of a sheared current profile on a curved
cylinder. Since the main focus was to gain an understanding of
the fundamental mechanisms of vortex shedding past curved
cylinders, a low Reynolds number, i.e. 100, was chosen.
Moreover, at this low Reynolds number the detailed flow
structures can be captured more clearly. In future computations,
the Reynolds number will be increased so that the development
of intrinsic three-dimensionality in the wake can be studied in a
progressive manner.
a) Quarter turn in uniform normal flow
The flow past a quarter turn with an aspect ratio of 12.5,
defined as the ratio of the radius of rotation from the centre of
curvature by the cross-section diameter of the cylinder, subject
to uniform flow normal to the plane of curvature has been
simulated. The diameter of the cylinder was 1 unit. This aspect
ratio is in the correct order of magnitude and representative for
flexible catenary riser pipes. Steel catenary riser pipes, on the
other hand, exhibit much larger curvature ratios.
The hybrid mesh created for this computation comprised a
total of 7464 elements, out of which 790 were the prismatic
boundary layer elements.
Figure 1 depicts a cut in the computational domain in the
streamwise direction and exhibits symmetry about the diagonal
of the region. The flow direction was in the y-direction,
effectively into the page. The inlet and outlet planes are
essentially parallel to the page. In the streamwise direction, the
distance to the outlet was 15 diameters from the axis of the
body and the inlet distance was 5 diameters. Lines A and B

correspond to the plane boundaries having been assigned a

symmetry boundary condition whereas lines C and D
correspond to the plane boundaries assigned a uniform velocity
boundary condition, V = [0 1 0].

Figure 1 : Computational boundaries (not to scale)

The computation was performed originally with 3 modes
per edge of each element (P = 2), and was then restarted with 5
modes (P = 4) and finally with 7 modes (P = 6). When the
computation was performed with 7 modes, the mesh accounted
for 715,456 local degrees of freedom per variable. Blockage for
these computational cases was 4.5%.
Figure 2 depicts the force coefficients in the x-, y-, and zdirections. The x- and z- directions represent the horizontal and
vertical component of the transverse force respectively and the
y-direction is the direction of the drag. The x- and z- forces
come out to be of equal amplitude and frequency as expected
for an axisymmetric body with symmetry boundary conditions.
As Figure 2 demonstrates, the curved cylinder, unlike the
straight cylinder which exhibits a drag force with a frequency
equal to twice that of the lift, accounts for a second lower
amplitude peak in the drag force as an effect of curvature.
Moreover, the mean in the lift force coefficient is not zero as is
the case for a straight two-dimensional cylinder. Unlike the
straight two-dimensional cylinder, the two vortices shed per lift
cycle do not have to be of equal strength. It should also be
mentioned, as an indication of the resolution of the mesh, that
when the polynomial order for the computation was increased
from 4 to 6, the lower amplitude peak in the drag force did not
exhibit any changes.

Copyright #### by ASME

axis of the vortex was found to be parallel to the axis of the

cylinder as Figure 4(b) depicts. The ring, being an
axisymmetric body, should have indeed been expected to shed
in phase along the span at this Reynolds number. In-phase
shedding at this Reynolds number would also be expected from
a straight cylinder implying that a sectional analysis might still
be appropriate for these flow conditions. Figure 4(a) shows the
extraction of pressure isocontours at a level of 0.30u2, which
corresponds to a base pressure coefficient of 0.82. Figure 4(b),
on the other hand, shows the extraction of pressure isocontours
at a level of 0.40u2, which corresponds to a base pressure
coefficient of 1.02, and their projection to the xy-plane. As
Figure 5 depicts, this pressure level is geometrically included in
the previous one as it corresponds to higher suction.

Figure 2 : Quarter turn in uniform flow normal to the plane of

curvature, Red = 100
The power spectra density for the forces in x-, y- and zdirections are shown in Figure 3. The Strouhal frequency for a
quarter turn in uniform normal flow has been found to be equal
to 0.175, a value larger than the documented Strouhal
frequency (0.167) for a straight two-dimensional cylinder at a
Reynolds number of 100. The drag force exhibits two
frequencies, 0.175 and 0.350. The latter is equal to twice the
Strouhal frequency and they are both of approximately the
same amplitude in the power spectra.

Figure 3 : Power spectra density for (a) Fx ; (b) Fy ; (c) Fz

Extraction of pressure contours at slices normal to the axis
of the curved cylinder in variable spanwise locations indicated
in-phase shedding as can be observed from Figure 2. Extraction
of pressure isocontours in three-dimensions verified this as the

Figure 4 : Pressure isocontours for a quarter turn in uniform

normal flow, Red = 100; (a) p = 0.30u2, Cpb = 0.82;
(b) p = 0.40u2, Cpb = 1.02, projection in the xy-plane

Figure 5 : Pressure contours for a quarter turn in uniform

normal flow, Red = 100; p = 0.30u2and p = 0.40u2
The root mean square (rms) value of the fluctuating lift
coefficient was found to be equal to 0.160.01. Graham [7]
compiled results produced from a number of two-dimensional
codes on the comparison of fluctuating lift coefficients for a
circular cylinder at a Reynolds number of 100. Bearman [6]

Copyright #### by ASME

presented them in a table. The CLrms values covered a range

from 0.16 to 0.34 and CLrms was found to be very sensitive to
detailed aspects of the simulations [6]. Specifically, Sherwin
[6] using the same spectral/hp element method as in this work
computed the CLrms value for the straight two-dimensional
cylinder at a Re of 100 to be equal to 0.24. Hence, the rms
value computed for the fluctuating lift coefficient of a quarter
turn at a Reynolds number of 100 in the present work is smaller
than that computed for a straight two-dimensional cylinder
upon employment of the same computational method.
Moreover, this value is in fact equal to the lowest value in the
range of rms lift coefficients for a straight two-dimensional
cylinder, an effect that could well be attributed to curvature.
b) Curved riser in contact with the seabed, sheared normal flow
Following the same procedure as for case (a), a quarter
turn with a vertical extension of two diameters ending semiburied in the seabed and subject to sheared normal flow was
also considered. This structure had the same aspect ratio, 12.5,
as before and the computational domain comprised of 7902
elements, out of which 864 were the prismatic surface
elements.
Artificial blending had to be applied at the touch down
point of the riser with the seabed to alleviate mesh generation
issues. Nevertheless, the resultant geometry has been
effectively an even closer representation of a catenary riser
being semi-buried in the seabed having seabed material adhered
to it in the long run of its lifetime in a well.
The bottom plane of the computational domain was
represented with a viscous wall as a boundary condition and an
exponentially decaying sheared profile was chosen for the
computational planes assigned a velocity boundary condition.
The shear profile applied was:

v = 1-e-(z+12.5)

This shear profile resembles that of a boundary layer

developing from the seabed. The velocity is equal to one at the
top of the riser and equal to zero at the bottom to conform with
the wall boundary. The shear profile is such that at 4.6
diameters from the seabed, the velocity achieves 99% of its
maximum value.
The computation was performed originally with 3 modes
per edge of each element (P=2) and was then restarted at P=4
and P=6. When a computation was performed at sixth
polynomial order, the mesh accounted for 760,536 local
degrees of freedom per variable. Blockage for these
computational cases was roughly in the order of 6%.
Figure 6 depicts the force coefficients in the x-, y-, and zdirections along with the pressure contours at slices normal to
the cylinder axis. The x- and z- directions represent the
horizontal and vertical component of the transverse force
respectively and y- is the direction of the drag. The force plots
indicate a modulation in the horizontal and vertical component
of the transverse force. Moreover, unlike case(a), the extraction
of pressure contours at the slices demonstrates lack of
correlation along the span of the riser pipe.

Figure 6 : Curved riser in contact with the seabed.

Shear flow normal to the plane of curvature, Red = 100
Moreover, in order to investigate the wake structures
developed by such a type of flow the 2 criterion presented by
Jeong & Hussain [2] was used for identifying the vortex cores
in the wake. The 2 criterion stems from the pressure minimum
criterion and dictates that the second largest eigenvalue of
S2+2 (where S is the rate of strain tensor and is the vorticity
tensor) must be negative in order to locate a vortex core. As
Figure 7 demonstrates, the vortex cores are no more parallel to
the cylinder axis.
In summary, the vortex shedding pattern past this structure
exhibited a lack of correlation along the cylinder axis
associated with a lower frequency modulation developing in the
horizontal and vertical directions of the transverse force and an
indication of a potential dislocation occurring as the curved
pipe approached the seabed.

Copyright #### by ASME

Figure 7 : Vortex cores in the wake of a curved riser in contact

with the seabed. Shear normal flow, Red = 100, 2 = -0.1

c) Curved riser in contact with an inviscid wall, sheared

normal flow
A direct comparison between cases (a) and (b) involves the
effects of both the sheared profile and the wall boundary.
Intermediate cases were therefore computed whereby the
bottom wall was assigned a symmetry boundary condition and
thus the effect of the sheared profile alone could be investigated
separately from that of the viscous wall. The bottom wall is
effectively now an inviscid wall and is therefore acting merely
as a plane blocking the flow, but not as a viscous boundary.
The mesh for this computational case is exactly the same
as the one for case (b) with the same sheared velocity profile
being applied. The force coefficients and the pressure contours
are depicted in Figure 8 and a preliminary comparison with
those for case (b), as depicted in Figure 6, can be drawn. The
force coefficients in general are approximately of the same
magnitude as those for case (b). The imposition of a different
wall boundary condition, however, has weakened significantly
the lower frequency modulation in the horizontal direction of
the transverse force, but not in the vertical one. In other words,
a sheared profile with a viscous wall exhibits a lower frequency
modulation in both the horizontal and vertical directions of the
transverse force whereas a sheared profile with an inviscid wall
exhibits a much weaker frequency modulation in the horizontal
direction of the transverse force than in the vertical one. In
addition, in the vertical component of the transverse force the
deviations from the mean at the beginning/end of each cycle are
much smaller and less regular for this case as opposed to case
(b).

Figure 9. The vortex cores are indeed not parallel to the

cylinder axis as has been the case for the sheared profile with a
viscous wall. This emphasises the significance of the sheared
profile versus the wall boundary. Nevertheless, comparing
Figures 7 and 9, one can claim as a preliminary comparison that
in the very near wake the difference between the viscous and
the inviscid wall consists of the immediate vortex tube ending
on the cylinder for the viscous wall as opposed to the inviscid
wall case where linking of the vortex structures occurs close to
the cylinder. In general, the shedding pattern for the inviscid
wall case seems to be more complex and the wake more
disorganised than that for the viscous wall.

Figure 9 : Vortex cores in the wake of a curved riser in contact

with an inviscid wall. Shear normal flow, Red = 100, 2 = -0.1
Hence, the significant differences between cases (a) and
(b) can be attributed more to the sheared profile rather than to
the wall boundary.

d) Curved riser in contact with an inviscid wall, uniform cross

flow

Figure 8 : Curved riser in contact with an inviscid wall.

Shear flow normal to the plane of curvature, Red = 100
In order to investigate the wake structures developed past
this type of flow, vortex cores at a 2 value of 0.1 were
extracted following the same criterion and are depicted in

In order to validate the above argument, a computation was

performed with exactly the same mesh as for cases (b) and (c),
but now a uniform velocity profile has been applied in the
normal to the plane of curvature direction. The wall boundary
condition remained a symmetry boundary condition
representing an inviscid wall.
The force coefficients for a segment of the force history
and pressure contours are depicted in Figure 10. The lower
frequency modulation in the horizontal direction of the
transverse force has now completely vanished. Moreover, the
drag force pattern depicted in Figure 10 exhibits noticeable
differences with the drag force coefficients for cases (b) and
(c).
At this point, however, it should be emphasised that this
case brings us back to case (a), which represented the base case,
for comparison. Even though there is a difference in the exact
geometry studied (this case ends semi-buried at the seabed as
opposed to case (a)), both cases had the same uniform normal

Copyright #### by ASME

flow velocity profile being applied. Specifically, the lower

amplitude peak in the drag direction for the quarter turn case
discussed as an effect of curvature is now apparent for this case
too. However, unlike the case for the quarter turn, there is a low
frequency variation. Note that this lower amplitude peak
phenomenon attributed to the unequal strength of the two
vortices did not become clear for cases (b) and (c) where the
effect of the sheared velocity profile was dominant.

Figure 11 : Vortex cores in the wake of a curved riser in contact

with an inviscid wall. Uniform normal flow,
Red = 100, 2 = -0.1

Figure 10 : Curved riser in contact with an inviscid wall.

Uniform flow normal to the plane of curvature, Red = 100
Finally, one can observe going from case (c) to the present
configuration, that the modulation in the vertical direction of
the transverse force has not completely vanished as in the
horizontal one. Overall, therefore, one can claim that going
progressively from case (b) to (c) and then (d), the horizontal
direction is affected before the vertical direction of the
transverse force.
Vortex core extraction in the wake of this flow is depicted
in Figure 11 employing the same 2 criterion. Contrary to cases
(b) and (c), the vortex cores are not inclined to the cylinder axis
as the curved pipe approaches the seabed. This verifies the
argument developed at the end of the section for case (c) i.e.
that the inclination of the vortex cores with respect to the
cylinder axis should be attributed to the sheared flow profile
primarily rather than to the wall boundary.

e) Quarter turn with an in-line extension in uniform flow

parallel to the plane of curvature
The flow past a quarter turn with an in-line extension of
ten diameters subject to uniform flow in parallel direction to the
plane of curvature was also computed. The mesh comprised of
10250 elements, out of which 1184 were the prismatic surface
elements. The computation was performed originally at second
polynomial order and was then restarted at fourth and sixth
order. When the computation was performed at sixth
polynomial order, the mesh accounted for 993,608 local
degrees of freedom per variable. Blockage for these
computational cases was 7.2%.
Figure 12 depicts the force coefficients in the x-, y-, and zdirections for a segment of the force history and the pressure
contours at slices normal to the cylinder axis. The x-direction is
now the direction of the drag and the y- and z- directions
represent the horizontal and vertical component of the
transverse force respectively. As the pipe bends to become
aligned with the free-stream, this structure is found to shed
much less vigorously and the flow does not stagnate.

Copyright #### by ASME

Figure 13 : Force history with interpolating polynomials of

order 4 and 6

Figure 12 : Quarter turn with an in-line extension in uniform

flow parallel to the plane of curvature, Red = 100
Even though Cz from Figure 12 may seem to exhibit an
increase, one must look at the whole force history. Figure 13
depicts the convergence to an almost saturated state achieved
by increasing the order of the interpolating polynomials from 4
to 6. The force coefficients depicted in Figure 12 correspond to
a segment towards the end of the force history depicted in
Figure 13 and represent a polynomial order of 6. The transition
from order 4 to order 6 occurs at time t = 350.5. Moreover, one
should note that the fluctuations and the scaling in the axes of
the x- and z- directions are significantly smaller compared to
those in the y- direction.
As a general comment, one should say that the force
coefficients produced for this case behave in a similar manner
as those for the standard example of a straight two-dimensional
cylinder. As Figure 12 depicts, the deviations from the mean
are much smaller for the drag force (x-direction) than for the
horizontal component (y-direction) of the transverse force. The
frequency of the drag was found to be equal to 0.35 and that of
the horizontal component of the transverse force to be equal to
0.175; hence the similarity with the two-dimensional cylinder
which exhibits a drag force with a frequency equal to twice that
of the lift force.
Furthermore, in order to investigate the wake structures
developed by such a flow the 2 criterion was employed again.
As Figure 14 shows, in the near wake the vortex cores are
straight, with their axes normal to the free-stream, but start to
bend further downstream in the wake.
Figure 15 depicts a projection of the vortex cores in the xyplane showing a staggered array of vortices in the wake.

Figure 14 : Vortex cores in the wake of a quarter turn with an

in-line extension in uniform parallel flow, Red = 100, 2 = -0.1

Figure 15: Projection of the vortex cores in the xy-plane

Experimental work was also conducted in the Hydrolab
facility of the Department of Aeronautics of Imperial College,
in order to compare the computational results with the flow
visualisations. A model riser with the same aspect ratio as in
the computational cases was towed in a tank [10] at an effective
Reynolds number equal to that of the computations. The flow
visualisation images obtained by illuminating dye diffused from
the stagnation line of the riser have shown good qualitative
agreement with the numerical investigations. Specifically, the

Copyright #### by ASME

flow visualisation image obtained for the wake of case (e) is

shown in Figure 16 and can be compared with Figure 14. The
bending of the vortices further downstream in the wake is more
apparent in the experiments than in the computations where the
mesh is less refined far away from the curved pipe.

quarter turn with a horizontal extension of ten diameters under

the following conditions:
in contact with the seabed; sheared flow parallel to the plane
of curvature
in contact with an inviscid wall; sheared flow parallel to the
plane of curvature
in contact with an inviscid wall; uniform flow parallel to the
plane of curvature
ACKNOWLEDGMENTS
Computational resources were provided by the Imperial
College Parallel Computing facilities (Fujitsu AP3000) and the
Computer Services for Academic Research (CSAR) at the
University of Manchester under the EPSRC grant number
GR/M08424.

Figure 16 : Flow visualisation for a quarter turn with an in-line

extension in uniform flow parallel to the plane of curvature, Red
= 100
CONCLUSIONS
This paper has presented the fluid dynamic loading and the
wake structures developed past curved cylinder flows either in
uniform and shear flow or in contact with the solid boundary in
directions normal or parallel to the plane of curvature.
The simulations in the direction normal to the plane of
curvature aimed at the comparison between the uniform flow
past a quarter turn and the more representative case for a
catenary riser, i.e. that of a deformed quarter turn with a
vertical addition of two diameters ending semi-buried in the
seabed and subject to sheared normal flow. The intermediate
stages were chosen so that the effects of the sheared profile and
the type of the wall boundary could be studied separately.
In the normal flow computations the effect of the shear
velocity profile has been found to be predominant when
compared to that of the wall boundary. Furthermore, the
experimental results for the flow in the direction parallel to the
plane of curvature have shown good qualitative agreement with
the numerical investigations. The force coefficients produced
for this case have been found to behave in a similar manner as
those for the standard case of a straight two-dimensional
cylinder.
Similarly to the study performed for the normal flow cases,
the effect of applying different velocity profiles and types of
wall boundary conditions on curved cylinder flows needs to be
investigated in the direction parallel to the plane of curvature.
Current work therefore includes computations of the flow past a

REFERENCES
[1]. Karniadakis, G. E. & Sherwin, S. J. (1999). Spectral/hp
Element Methods for CFD. Oxford University Press.
[2]. Jeong, J. & Hussain, F. (1995). On the identification of a
vortex. J. Fluid Mech 285, 69-94.
[3]. Peiro, J., & Sayma, A. I. (1995). A 3-D unstructured
multigrid Navier-Stokes solver. In K. W. Morton &
M.J.Baines, eds., Numerical Methods for Fluid Dynamics,
V, Oxford University Press.
[4]. Peraire, J., Peiro, J., & Morgan, K. (1993). Multigrid
solution of the 3-D compressible Euler equations on
unstructured tetrahedral grids, Int. J. Numer. Methods
Eng. 36, 1029-1044.
[5]. Vandiver, J. K. & Li, L. (1994). SHEAR7 Program
Theoretical Manual. Department of Ocean Engineering,
MIT, Cambridge, MA, U.S.A.
[6]. Bearman, P. W. (1998). Developments in the
understanding of bluff body flows. JMSE International
Journal, Series B, Vol.41, No.1.
[7]. Graham, J. M. R. (1993). Comparing computation of flow
past circular cylinders with experimental data. In Bluff
body wakes, dynamics and instabilities. Springer Verlag,
Berlin, 317-324.
[8]. Karniadakis, G. E., Israeli, M., Orszag, S. A. (1991).
High-order splitting methods for the incompressible
Navier-Stokes equations. J. Comput. Phys 97, 414-443.
[9]. Owen, J. C. (2001). Passive Control of Vortex Shedding in
the Wake of Bluff Bodies. Ph.D. thesis, Imperial College
of Science, Technology and Medicine.
[10]. Sherwin, S. J. & Karniadakis, G. E. (1996). Tetrahedral
hp finite elements: Algorithms and flow solutions. Journal
of Computational Physics 124, 14-45.

Copyright #### by ASME