
\ .
M
All of the above equations apply to cases where axial flow induced by relative lateral motion of the
cylinders is insignificant either because the flow is blocked at the ends or because the cylinders are long
enough to ignore end effects. For common cases such as the downcomer annulus (See Figure  Features of
the AP1000 Structural Design), axial flow is blocked at one end open at other. For these situations it is
common to employ the following approximation that applies only to the thin annular case.
Figure 2. Shell Element Model of Core Shroud
Considering a finite length thin cylindrical annulus open at one end and imposing relative lateral
acceleration while permitting only axial flow results in an added axial hydrodynamic mass:
3
1
(5)
3
axial
h
a
M L
c
=
This axial mass effectively acts in series with the circumferential mass given previously so that the two
may be put together to give the hydrodynamic mass for the finite length case:
3
2
1
(6)
1 3
axial
finite
axial
h h
h
h h
M M
a L
M
M M c
a
L
 


=

+
 
 +


\ . \ .
The final beam mass matrix then has the form:
1
1
1 1 2
( )
[ ] (7)
( )
finite finite
finite
finite finite
h h
h h
M M M
M M M M M
+  
 =

+ + +
\ .
M
Equation (7) represents the total hydrodynamic mass for the beam mode. It is distributed to nodes over the
height of the model using either STIF27 or FLUID38 elements.
Added Mass Coupling for Shell Modes
The fluid displacements and resultant forces from pressure acting on the structures will be defined by a
Fourier series expansion. This defines a set of generalized forces, F
n
and F
n
,and generalized
displacements, A
n
and A
n
0
1 1
( )
( , ) ( ) cos( ) ( ) sin( ) (8)
2
r n n
n n
A t
U t A t n A t n
= =
= + +
&&
&& && &&
0
1 1
( )
( , ) ( ) cos( ) ( ) sin( ) (9)
2
r n n
n n
F t
F t F t n F t n
= =
= + +
Next, the shell mode mass matrices, [M
n
] can be defined as relating the generalized force and
accelerations:
[ ] (10)
a a
b
b
n n
n
n
n
F A
F
A
=
` `
)
)
M
&&
&&
[ ] (11)
a a
b
b
n n
n
n
n
F A
F
A
=
` `
)
)
M
&&
&&
where subscripts a and b are used for the inner and outer shells.
The fluid mass matrices are defined using solutions of the equations for incompressible flow whereby the
acoustic pressure distribution within the fluid annulus can be obtained using separation of variables. The
complete solution, given in Reference 5, expresses the radial pressure distribution using Bessel functions,
the circumferential distribution as the sine and cosine terms, and the axial distribution in form a Fourier
series. The full solution is somewhat complex but can be simplified by noting that for the range of
structural frequencies of interest, the axial length of the cylinders is small compared to the wavelength of
acoustic waves and the solution from Reference 5 can be written as:
2 2 2 1 1
2 2 1 1 2 2 2
2 2
2 2 2
( ) 2
[ ]
( ) 2 ( )
(1 ) 2
(12)
(1 ) 2 (1 )
n n n n
n
n n n n n n
n n
n n n
a a b a b L
n b a a b b a b
a r abr L
n r abr b r
+ +
+ +
  +
=

+
\ .
  +
=

+
\ .
M
where / r a b =
It is emphasized that use of this equation implies the same assumptions as used for the Fritz mass of FLUID
38:
The motions are small with respect to annulus width.
The fluid is incompressible (fluid velocities are less than 10% of speed of sound in the fluid.)
The axial length of the channel is small relative to wavelength of propagating disturbances.
To compare shell mode and beam mode masses it is helpful to expand the terms in equation 12 as follows:
2
2 2 4 4 6 6 8
2 2
1 1 1 1 1 2
(1 log ( ) log ( ) log ( ) [ ] (13)
3 45 945 1 log(1/ )
n
n
r
r n r n r n O n
n r r n
+
= + + +
and
2 2 4 4 6 6 8
2 2
1 1 1 7 31
(1 log ( ) log ( ) log ( ) [ ] (14)
6 360 15120 1 2log(1/ )
n
n
r
r n r n r n O n
n r r n
= + +
Then, for 1/3 x log
2
(r) n
2
<< 1:
2
2 2
1 1 1
(15)
1 log(1/ )
n
n
r
n r r n
+
Similarly for equation 14:
2 2
1 1
(16)
1 2log(1/ )
n
n
r
n r r n
These approximations are sufficiently accurate for the cases of general interest, e.g. for r 0.8 the errors in
the approximations are less than {2%, 7%, 15%, 26%} for n= {1,2,3,4}.
Therefore reasonable approximation for thin fluid annuli is:
2
2 2
[ ] (17)
log( / )
n
a ab L
n b a ab b
 
==

\ .
M
This gives the very simple result that the shell mode mass matrices as defined above are equal to 1/n
2
times
the beam mode mass matrix.
The 1/n
2
result applies only to the circumferential part of the mass. For a finite length cylinder, added
mass for flow in the axial direction will be in proportion to the circumferential half wavelength,
= a /
n. Therefore equation 5 can be generalized for n 1 to be:
3
1
(18)
3
axial
h
a
M L
n c
=
The axial mass varies in proportion to 1/n versus the 1/n
2
behavior of the circumferential mass. Therefore,
since the axial mass acts in series with the circumferential mass, the adjustment for finite length has much
less significance for higher modes; nevertheless the axial length correction can still be approximated as
described in deriving equation 6.
Approach to Representing Shell Mass Matrix in ANSYS
The circumferential variation of radial displacements at any elevation of the cylindrical shell can be
expressed as a Fourier series given in equation 8.
Similar expressions can be written for tangential and vertical displacements but they will not be needed
since hydrodynamic forces are only radial.
The n=0 (or A
o
) term is the breathing mode; n=1 terms are the beam mode terms and n 2 terms are shell
modes. The cosine terms are symmetric about the =0 plane and the sine terms are antisymmetric. The
breathing modes for these shells have a very high frequency so that this term is not included in models
discussed later; however, it is included here for completeness. Making use of orthogonality, the Fourier
coefficients are:
2
0
1
( ) cos( ) d for n 0 (19)
n r
A U n
2
0
1
( ) sin( ) d for n 1 (20)
n r
A U n
If the shell is represented by N equally spaced (=2/N) nodes using N shell elements around the
circumference, the previous integrals may be approximated by the following sums:
1
2
cos( ) for n 0 (21)
i
N
n r i
i
A U n
N
=
=
1
2
sin( ) for n 1 (22)
i
N
n r i
i
A U n
N
=
=
where U
r
is the radial displacement.
Equations 21 and 22 can be entered into ANSYS as constraint equations. This is done by creating new
nodes which will be referred to as "Fourier" nodes and setting the displacement of these Fourier nodes to
the above relations, i.e.
1
2
cos( ) for n 0 (23)
n i
N
x r i
i
U U n
N
=
=
1
2
sin( ) for n 1 (24)
n i
N
y r i
i
U U n
N
=
=
For unequally spaced nodes, =(
i+1

i1
)/2 , and equations 21 and 22 become:
1 1
1
1
cos( ) for n 0 (25)
2
i
N
i i
n r i
i
A U n
+
=
1 1
1
1
sin( ) for n 1 (26)
2
i
N
i i
n r i
i
A U n
+
=
Equations 23 and 24 are modified similarly. These are the equations used for the noncircular and irregular
mesh of the core shroud.
Typically, the n=0 mode is not needed and no more than the beam and first 3 shell modes are significant.
Therefore, 4 nodes (at each shell elevation) are created each with 2 degrees of freedom (U
x
and U
y
). Note
that the degrees of freedom at the new nodes are in the Cartesian x and y directions. The z direction is
assumed to be the axis of the cylinders.
Equations 23 and 24 are in a form that ANSYS defines as linear constraint equations. If these constraints
are entered into ANSYS, the displacements calculated at the "Fourier" nodes will be equal to the
coefficients of the Fourier expansion of radial displacements. For the case of n=1, these coefficients are
equal to the beam mode displacements in the Cartesian x and y directions.
Fourier nodes are created on both the inner shell and outer shell. The hydrodynamic mass is entered as
STIF27 mass matrices connecting the Fourier nodes ( Note: The case that either the inner or the outer
cylinder is modeled as a beam is discussed below.)
1 1
1 1 2
( )
(27)
( )
n n
n n
a a
x x
n n hn
b b
n hn n n hn x x
F U
M M M
M M M M M F U
+  
=
` ` 
+ + +
\ .
) )
&&
&&
1 1
1 1 2
( )
(28)
( )
n n
n n
a a
y y
n n hn
b b
n hn n n hn y y
F U
M M M
M M M M M F U
+  
=
` ` 
+ + +
\ .
) )
&&
&&
where a and b correspond to Fourier nodes of the inner and outer shell respectively.
Note that the mass matrices will be different for each mode, but it is assumed that they can all be written in
this form. The mass matrix for n=1 will be taken as the usual Fritz mass. As discussed previously an initial
estimate mass for n2 can be obtained by multiplying the masses for n=1 by 1/n
2
, i.e. use 1/4, 1/9, and 1/16
for modes 2, 3, and 4.
The reactor vessel is modeled as a beam which is equivalent to assuming that its shell modes are "rigid".
Therefore the mass matrices couple the reactor vessel and core barrel for n=1 only. For n=1 the mass matrix
is placed between the Fourier node of the core barrel and the node of the beam element. For n>1, the mass
matrix is between the Fourier node of the core barrel and a fictitious fixed (grounded) node.
Added Mass for Noncircular Fluid Annuli
The problem of added mass for noncircular annuli has been studied experimentally in Reference 4.
Additional unreported studies have been performed using models that make used of 3D fluid elements.
Nevertheless this remains an area in which further research is needed.
In the present study, the beam mode mass for noncircular annuli, such as that between the core shroud and
core barrel, was determined by using an equivalent circular radius. The equivalent circular radius was taken
to be that which gives the same fluid cross sectional area as the actual geometry. Experimental
measurements have shown this to be reasonable for the beam mode. For higher modes the 1/n
2
adjustment
to the beam mode mass was used. Although for these structures this produced reasonable agreement with
available data and models using 3dimensional fluid elements, this should not be used as a general rule for
noncircular annuli without further study or experimental data.
Procedure
In summary the methodology is applied as the following steps:
a) Create "Fourier" nodes at each elevation of each shell. If the first m shell modes are
considered, there will be m separate nodes per each shell per each elevation at which the
hydrodynamic mass is to be distributed to 2m Fourier degrees of freedom. Note that nodal
coordinates on the shells are rotated to the cylindrical coordinate system while Fourier nodes
remain in a Cartesian nodal coordinate system.
b) Enter constraint equations to define the Fourier node displacements. For each Fourier node
there will be an "x" and "y" direction constraint. Terms of the constraint equation are easily
entered in a macro with doloops over elevation, direction and shell mode.
c) Enter the 2x2 hydrodynamic masses between adjacent (in elevation and direction) Fourier
nodes.
It is recommended that:
1) First run the model for inair frequencies before adding the Fourier nodes and constraints.
Add the Fourier nodes and constraint equations but set mass matrices to small values. Verify
that the inair modes and frequencies are not affected by the constraints and Fourier nodes.
2) Add the hydrodynamic masses for the beam mode and verify that the beam mode is correct
and that inair shell modes are not affected.
3) Add masses for shell modes to complete problem.
4) Estimate the hydrodynamic mass using the 1/n
2
and 1/n rules discussed above. These values
should be treated as initial estimates with refinement based on experimental data.
Analysis
A full discussion of the AP1000 structural model and results is beyond the limits of this paper; however a
brief description of the model will explain the benefits of using the methods discussed above to obtain an
economical model.
The full model of the AP1000 system included over 15000 elements with over 85000 active degrees of
freedom (DOF). The majority of the elements and DOF were associated with the core shroud which had
over 11000 shell elements and 67000 DOF (See Figure Shell Element Model of Core Shroud). The core
shroud was represented as a superelement (MATRIX50) with 240 master DOF. The master DOF included
the Fourier nodes used for representing the hydrodynamic mass for the core shroud to core barrel annulus
for the beam and first 3 shell modes. Using the superelement, the number of active DOF in the analysis run
was reduced from approximately 85,000 to 19,000.
In general the analysis models included a small number (< 100) of nonlinear gap, contact and sliding
elements at supports of the vessels. Therefore most analyses of the response of the AP1000 structures due
to turbulent forces were performed using nonlinear transient dynamic analysis with applied pseudo
random forcetime histories. Following the timehistory analysis, the super element solution was expanded
at selected times. Finally, submodels were analyzed to calculate peak stresses in localized regions of the
core shroud and core barrel.
Conclusion
The analyses performed for the AP1000 plant demonstrated that the method of including hydrodynamic
mass for shell modes will reduce models to practical sizes needed for nonlinear transient dynamic analysis.
Had the same models been prepared using solid fluid elements, they would have been an order of
magnitude more costly to run if even possible. The method reported here has significant advantages in its
compatibility with using superelements to represent portions of the structure; whereas models with fluid
elements typically have coupling that makes division into superelements much more difficult.
For this and similar dynamic analyses of fluidstructure interaction problems, (Reference 6) ANSYS
constraint equations have been demonstrated to be a useful means to minimize the number of element that
would be needed to represent coupling of incompressible fluids.
The extension of models to include shell modes is essential in more accurate assessment of flow induced
vibration (FIV). These assessments are necessary in that relative movements of structures that surround the
reactor core could lead to core damage or interference with control rod movement. Flowinduced
vibrations may also be a cause of some cases of loose parts and indication of other problems in the system.
Finally, although FIV is not generally the controlling loading for stress levels in these structures, the
assessment of flow induced vibration is required in evaluating normal operating stress and fatigue damage.
The initial application of the methodology described in this paper has been for assessment of structural
vibrations resulting from lateral forces due to turbulent flow; however, the models have also been used in
analysis of structural response due to vertical forces. The methods may also be used in future analyses for
other dynamic loadings including pumpinduced acoustic waves, seismic, and LossofCoolantAccident
(LOCA) analyses.
References
1) M. D. Snyder, et.al., AP1000 Reactor Internals FlowInduced Vibration Program, 2003
International Congress on Advances in Nuclear Power Plants (ICAPP 03), May 47, 2003,
Crdoba, SPAIN.
2) Yoshio Miyake and Hiroshi Mukai, For the Long Term Stable Supply of Electric Energy,
Mitsubishi Heavy Industries, Ltd. Technical Review, Vol. 40, No.1, February 2003.
3) R. J. Fritz, The Effect of Liquids on the Dynamic Motion of Immersed Solids, Journal of
Engineering for Industry, Transactions of the ASME, February 1972.
4) S. Nakura, et.al., Basic Test of Added Mass and Added Damping of a Rigid Body Enclosed in a
container in Narrow Passage, Proceedings of ICONE 8, 8th International Conference on Nuclear
Engineering, Baltimore, MD., April 26,2000.
5) M. K .AuYang, Dynamics of Coupled FluidShells, Journal of Vibration, Acoustics, Stress, and
Reliability in Design, , Transactions of the ASME, Vol. 108, July 1986.
6) R.E. Schwirian, D.R. Forsyth, M. D. Snyder, et.al., "The BaffleBarrelBolting Analysis
Program: Evolution & Technical Accomplishments", Ninth International Conference on Nuclear
Engineering , ICONE9, Paris, April, 2001.