Session 8 Basic Theory & Roll Damping – Intact and Damaged Ships

Some Results from a New Time-Domain Bilge Keel Force Model

by D. S. Greeley
David S. Greeley
Applied Physical Sciences Corporation
ABSTRACT
A new non-linear, time domain bilge keel force model was recently developed for inclusion in the new
time-domain seakeeping/maneuvering in waves code TEMPEST, being developed by NSWCCD. This
bilge keel force model combines a full unsteady extension of Bollay’s non-linear low aspect ratio lifting
surface theory for cases with adequate forward speed with a more conventional approach for cases with
zero or low forward speed, using Morison’s equation. This paper presents some representative results
from the new bilge keel force model for a surface combatant for various roll amplitudes, roll periods, and
forward speeds.
KEYWORDS
Bilge-keel forces; time-domain seakeeping ; unsteady lifting surface theory

curve fits of free roll decay data from a still water

INTRODUCTION model test (either stationary or with forward
The calculation of forces on ship bilge keels speed) in an attempt to discern new “physics” of
continues to be a challenging but important aspect rolling in a seaway, when this relative motion
of ship hydrodynamics, because of the large effect situation is ignored. A thorough analysis may
of these forces on rolling, and the consequent someday indicate exactly under what conditions
possibility of capsizing and loss of the ship. In the relative motion problem may be ignored and
the past, the estimation of these forces for ship roll decay data obtained from still water roll decay
motion simulations has traditionally relied on data constitutes the “right” input to a seakeeping
experimental data, and semi-empirical techniques code, but to the author’s knowledge no such
derived largely from experimental data (Himeno analysis has yet been done. The relative motion is
1981). More recently, an assault on this problem largely ignored today because the tools to handle
using modern computational techniques has begun it are just being developed.
to bear fruit. Before getting to the results which
2) Ship motion researchers commonly refer
are central to this paper, it is worthwhile to review
to possible hydrodyamic “memory” in a
some “big-picture” aspects of the forces on bilge
seakeeping problem as being associated only with
keels:
free surface waves. However, unsteady shed
1) Except for some exceptional cases such vorticity (associated with time-varying
as yaw or rudder-induced rolling, ships roll circulation) also constitutes a powerful
because of the excitation from incident waves in mechanism for hydrodynamic “memory”. In the
the ocean. It is obvious from first principles that case of bilge keels, if it is assumed that only
the forces on the bilge keels are not dependent “linear” lift is generated (where the trailing
strictly on the roll and roll rate (or even vorticity is shed only from the trailing edge), this
complicated functions of these variables alone!), shed vorticity will have minimal impact.
but rather on the relative motion between the bilge However, numerous URANS calculations and
keels and the water. This relative motion includes PIV measurements (Irvine 2006) have shown that
all motions of the ship, the forward speed of the for low aspect ratio bilge keels, trailing vorticity is
ship, the incident wave orbital velocities, and the shed from the entire side edge of the bilge keel,
velocities associated with the wave diffraction, the and this shed vorticity has a large impact on the
body radiation, and the steady waves generated by forces developed on the middle and aft parts of
the ship. It makes little sense to get completely the bilge keel. Bilge keel force models for zero or
caught up with an elaborate analysis and fancy low forward speed that use the Morison equation

283
 Proceedings of the 12th International Ship Stability Workshop

implicitly capture this memory effect, in that the techniques: a) high Reynods numbers, b) low
data used in the Morison equation depends on the Mach numbers, and c) sharp edges which fix flow
history of the flow through the Keulegan- separation locations. It is the author’s opinion
Carpenter parameter. For the situation with that the community would be better served if
forward speed, either URANS calculations or an “lifting forces” were clearly differentiated from
unsteady lifting surface theory that explicitly deals true “viscous” forces.
with side edge vortex shedding and the correct 5) For large roll angles, which are the most
accounting for unsteady shed vorticity is important to consider when considering ship
necessary to capture this memory effect. The safety, bilge keel interactions with the free surface
memory is dependent on all of the past motions of are a distinct possibilty, even to extent of partial
the ship and the incident waves, not just roll or complete bilge keel emergence. It is almost
motion. certainly more important to try to account for this
3) It is common to recognize that bilge keels effect, at least approximately, than to get into
contribute to the added roll moment of inertia of protracted arguments about the correct functional
the ship. It is also common to assume that this form of the bilge keel damping for high roll
contribution is constant and may be easily angles so that low roll angle data may be
estimated from the added mass of a 2D flat plate. extrapolated to high roll angles.
However, if one uses a zero speed bilge keel force 6) Many of the considerations above argue
model based on the Morison equation, then the for a time-domain treatment of the bilge keel force
effect of shed vorticity on the bilge keel added problem, in order to handle the inherent non-
mass (ship roll moment of inertia) is implicitly linearities and memory effects. Much current
included because of the dependence of the inertia effort, including the author’s, is in this direction.
coefficient in the Morison equation on the However, recognizing the usefulness of
Keulegan-Carpenter parameter. In the case with frequency-domain seakeeping calculations, it
forward speed, the situation is more complicated would be worthwhile to see if current and future
and the added moment of inertia of the bilge advancements in time-domain calculations could
keels, like the lifting forces which contribute to be transferred to the frequency domain, perhaps
roll damping, are in fact dependent on the history using harmonic balance techniques.
of the motion because of the unsteady shed
vorticity.
NEW BILGE KEEL FORCE MODEL
4) It is typical within the ship motion
community to refer to any force that cannot be In (Greeley 2010a) a new time-domain bilge keel
modeled in a potential flow sense with a single- force model was presented, which was developed
valued velocity potential as a “viscous” force. to be incorporated into the US Navy’s new
This implies, especially to newcomers to the field, seakeeping / maneuvering in waves code
that these forces can only be addressed in the TEMPEST (Belknap 2010). The main constraint
context of RANS or URANS calculations. on this new force model, besides the obvious
Obvious examples are the side force and yaw requirement for maximum accuracy, was that it
moment on a yawed ship hull, and the forces on execute quickly so that the final TEMPEST code
bilge keels. This is tantamount to ignoring all of could run in near real-time. This obviously meant
the excellent work for calculating lifting potential that URANS approaches could not be considered.
flows about arbitary bodies done by John Hess In the end, a hybrid time-domain force model was
and others, starting in the 1970’s (Hess 1972), developed that is based on the relative motion
which became the main aerodynamic design tool between the bilge keel and the water (as described
for airplanes until Euler and RANS techniques above), and consists of two components:
supplanted them because these new techniques 1) For zero or low forward speeds, where
could deal with Mach numbers approaching or the unsteady angle of attack of the bilge keels
exceeding 1. In fact, many (but certainly not all) exceeds 45 degrees, the force model uses the
situations in ship hydrodynamics are perfect for Morison equation with the empirical database for
the application of lifting potential flow flat plates in unsteady flow presented by Sarpkaya

284
 Session 8 Basic Theory & Roll Damping – Intact and Damaged Ships

(2006). Two different techniques were developed corresponding CFD results, done using CFDShip-
and evaluated for determing the Keulegan- Iowa. These results were presented at model scale
Carpenter parameter from the relative motion so that the experiments and CFD computations
history, which is a necessary input to the could be compared directly; we have done our
Morisson equation database. example computations at ship scale, but we
2) For those situations with forward speed present the results in terms of non-dimensonal
where the angle of attack of the bilge keel is 45 coefficients so that model and full size results may
degrees or less, Bollay’s theory (Bollay 1936) for be compared directly. The major results that we
low aspect ratio wings was extended for unsteady will use for comparison are roll decay coefficient
flow with an arbitrary distribution (in time and versus average roll angle and roll period versus
space) of normal and tangential onset velocities to cycle number, as shown in their Figure 9
the bilge keel. The vorticity, which is assumed to (reproduced below as Figure 1).
be continuously shed from the side edge (as well
as the trailing edge) of the bilge keel according to
Bollay’s theory, is rigorously accounted for in an
unsteady lifting surface fashion, and the full non-
linear Bernoulli equation is used to compute the
forces on the bilge keel. This model has been
demonstrated to closely replicate URANS force
results, even at an unsteady angle of attack of 24
degrees.
The final bilge keel model (Greeley 2010b)
contains the above two modules, all appropriate
logic to switch between the force modules as
appropriate, and a model for the unsteady pressure
forces (due to bilge keel action) acting on the hull
surface (which uses the exact shape of the hull
adjacent to the bilge keels). In addition, the final
model includes an approximate but physically-
based model for the change in bilge keel forces as
a bilge keel emerges from the water, and a model
for the slam forces on the bilge keel as it re-enters
the water.
This new bilge keel force model is currently
being implemented into TEMPEST by the Fig. 1: Decay Coefficient and Roll Period for DTMB
TEMPEST team, so no results are yet available Model 5415 (from Miller 2008) (scale ratio = 24.84)
from the new bilge keel model actually used with
TEMPEST. We present below some computations The excellent match between CFD and
using the new bilge keel force model for a ship experiment (at the higher roll angles) allows one
rolling in calm water. to have confidence in the contributions of the
bilge keels to the motions, as determined by the
difference in CFD computed motions with and
EXAMPLE CALCULATIONS without bilge keels. From the data presented, we
The case chosen for the example calculations can pick off the difference in roll decay
presented here is the U.S Navy pre-contract DDG- coefficient due to the bilge keels, and the
51 hull form, as represented by NSWCCD model difference in roll period due to the bilge keels.
5415. This hull form has been widely studied For convenience, these values have been
around the world. In particular, Miller (2008) has converted to equivalent linear damping
presented calm water roll decay measurements (at coefficients b44 and added moment of inertia
Fr=0.0 and Fr=0.280) for Model 5415 and coefficients a44, and we present here their non-

285
 Proceedings of the 12th International Ship Stability Workshop

dimensional forms as recommended by Himeno the last complete roll cycle in the calculations.
(1981): Both the normal pressures on the bilge keel and
a the computed pressures acting on the nearby hull
a44 '  44 2 (9) were used to compute the bilge keel roll moments,
B
to be consistent with the previously mentioned
CFD computations. A typical plot of the
b44 computed roll moments (for the starboard bilge
b44 '  (10) keel only) is shown below.
2g
B 2

where B is the beam of the ship, and g is the

acceleration due to gravity. These coefficients are
terms in the single degree of freedom equation for
the roll of the ship:

( I 44  a44 )
  b44   c44   M (t ) (11)

The inflow to the bilge keel for our example

calculations is determined by using a lifting Fig. 2: Computed Starboard Bilge Keel Roll Moments at
potential flow model, following the work of Hess Froude No. = 0.280 (full scale)
(1972). The hull surface is paneled using source
panels to enforce the zero normal velocity Figure 2 shows the roll angle, the bilge keel roll
boundary condition on the ship hull, and an moment from the unsteady lifting surface
interior vortex lattice system is used to represent subroutine (mxuls3) which is the correct roll
the lifting action of the sonar dome and hull. A moment to use in this case, the corresponding roll
non-linear, equal pressure Kutta condition is moment if just the zero-forward-speed Morison
applied at the trailing edge of the sonar dome and equation were used (mxme), as well as the
skeg in order to determine the strengths of the computed roll moment due to bilge-keel induced
interior vortex lattice system. Because the rolling pressures on the hull (hmx). As is well known,
motion considered is of relatively high frequency, the bilge keel forces decrease with forward speed,
the best approximation of the free surface so the unsteady lifting surface computation shows
available in this potential flow calculation is that a smaller roll moment at this medium Froude
the calm water surface be considered a plane of number than the zero-forward-speed Morison
zero perturbation potential (Newman 1977): this equation would indicate.
may be realized by using a negative image above The computed roll moments versus time
the waterplane for all of the singularities were then least squares fitted with a model with 4
representing the hull. terms: one for the added moment of inertia, and 3
The bilge keel calculations for each for the roll damping moment assuming that the
operating condition were done by computing the roll damping can be expressed in the following
starboard bilge keel inflow (both tangential and form (Himeno 1981):
normal velocity distributions along the bilge keel) b ( )  b   b    b  3
44 1 2 3 (12)
for a series of roll angles and roll velocities
corresponding to one sinusoidal roll cycle, using The equivalent linear damping for this Froude
these velocities to “drive” the new bilge keel force number, roll frequency, and roll angle amplitude
model for several cycles to ensure that the is then computed from these three components
hydrodynamic memory effects associated with the and used to compute the final b44:
shed vorticity had reached a steady sinusoidal
pattern, and then analyzing the force results for

286
 Session 8 Basic Theory & Roll Damping – Intact and Damaged Ships

8 3 in the hull potential flow computations). Still, the

be ( Fr , ,  A )  b1   Ab2  2 2Ab3 (13) reasonable agreement for the bilge keel damping
3 4
values is encouraging.
The most unusual feature of the predicted
b44  be ( Fr , ,  A ) (14) damping due to the bilge keels shown in Figure 3
Figure 3 shows the computed bilge keel is the large increase in damping expected at
contribution to the added moment of inertia and Fr=0.07. This comes right out of the unsteady
equivalent linear damping coefficients for the lifting surface model. This lifting surface model,
DDG-51 hull form at various Froude numbers, for since it only responds to the normal and tangential
a roll amplitude of 15 degrees and a full scale roll velocities supplied to it, does not care about the
period of 11.2 seconds. The added moments of actual Froude number. Rather, this Froude
inertia due to the bilge keels inferred from the number of 0.07 with a roll amplitude of 15
data of Miller (2008) are two to three times the degrees and roll period of 11.2 seconds happens
values shown in Figure 3. The reason for this to correspond to an unsteady angle of attack and
large discrepancy is unknown at present – it may reduced frequency combination at which the
be related to fact that during the experimental vorticity shed from the side edges of the bilge keel
work and CFD computations, the model was on the forward third of the bilge keel has a huge
forced to oscillate about its center of gravity. impact on the load experienced on the after half of
With the coupling of roll and sway due to added the bilge keel – the biggest computed unsteady
mass terms, perhaps a two degree of freedom pressures on the bilge keel are in fact near 75%
model (roll and sway) is required to model the chord! This kind of hump in the bilge keel
forced roll experiment and the influence of the damping is not expected according to generally
bilge keels on the added moment of inertia in a used bilge keel force models or seen in most
meaningful way. In any event, this large model testing. In fact, there is sometimes a dip
discrepancy requires more study in the future. seen in the bilge keel damping in this Froude
number region. However, Irvine (2006) shows
large scale effects on bilge keel forces in this
Froude number range for geosims of Model 5415
tested at various facilities, with the smaller models
showing significantly less damping than the larger
models, so there are a number of issues that need
examing here: 1) re-examination and possible
additional testing and CFD runs for this Froude
number (reduced frequency) range, and 2)
comparison (if possible) between model scale
predicted roll damping and full scale ship roll
damping in this Froude number range to make
sure we understand the nature of possible scale
effects, and possible shortcomings in the unsteady
Fig. 3: Computed Bilge Keel Contribution to Added
Moment of Inertia and Equivalent Linear Damping
lifting surface model as currently implemented.
Additional computions with the new bilge
Two experimental/CFD data points for the keel model were done to look at variations in roll
non-dimensional bilge keel damping are also amplitude and roll period (for the full scale ship).
shown in Figure 3 – these were deduced from the Figure 4 shows the computed variation in a44 and
experimental / CFD results for the model shown b44 with roll period for a Froude number of 0.280
in Figure 1. A perfect match between the current and a roll amplitude of 15 degrees, while Figure 5
bilge keel model calculations and these points shows a similar variation with roll amplitude for
should not be expected, given the approximate Froude number=0.280 and a roll period of 11.2
nature of the bilge keel inflow calculations (not sec (full scale). As expected, since the roll
having a proper representation of the free surface damping actually has a large quadratic and cubic

287
 Proceedings of the 12th International Ship Stability Workshop

component, the equivalent linear damping is a center of gravity (regardless of added mass
strong function of roll angle. effects) also requires a re-examination.
ACKNOWLEDGEMENTS
The development of the bilge keel model referred
to in this paper was done initially with support
from NAVSEA SE/TA funds (James Webster,
Program Manager) and later support from ONR
(Dr. Pat Purtell, Program Manager). Dr. William
Belknap and Dr. Arthur Reed at NSWCCD were
the technical monitors for the development of this
bilge keel model. Simmy Willemann and Brian
Petersen of APS assisted with the calculations
shown in this paper. The opinions expressed here
are those of the author.
Fig. 4: Variation in Bilge Keel Parameters with Roll
Period (full scale) REFERENCES

Belknap, W.F. and Reed, A.M. (2010), “TEMPEST –

New Computationally Efficient Dynamic Stability
Prediction Tool”, 11th International Ship Stability
Workshop, Wageningen, June 2010.

Bollay, W. (1936), “A New Theory for Wings of Small

Aspect Ratio”, Caltech Ph.D. thesis, Pasadena,
CA, 1936.

Greeley, D.S. and Petersen, B.J (2010a), “Efficient

Time-Domain Comptation of Bilge Keel Forces”,
28th ONR Symposium on Naval Hydrodynamics,
Pasadena, CA, September 2010.

Greeley, D.S. and Pertersen, B.J.(2010b), “TEMPEST

Level III Bilge Keel Force Model Theory”,
Fig. 5: Variation of Bilge Keel Parameters with Maximum Applied Physical Sciences Report, September
Roll Angle (full scale) 2010.

CONCLUSIONS Hess, J.L. (1972), “Calculation of Potential Flow About

Some example calculations from a new time- Arbitrary Three-Dimensional Lifting Bodies”,
domain bilge keel force model have been McDonnell-Douglas Report No. MDC J5679-01,
October 1972.
presented and an attempt has been made to
compare the results from this model with existing Himeno, Y. (1981), “Prediction of Ship Roll Damping
experimental and CFD results for DTMB Model – State of the Art”, University of Michigan
5415. Some encouraging agreement with existing Department of Naval Architecture and Marine
experimental and CFD data has been obtained Engineering Report No. 239, September 1981.
even with approximate calculations of the bilge
keel inflow, but questions about the bilge keel Irvine, M., Atsavapranee, P., Carneal, J. et al (2006),
force behavior, especially possible scale effects “Comparisons of Free Roll Decay Tests for Model
near a Froude number of 0.07, require further DTMB 5415/2340/5512, and Investigation of
investigation. The interpretation of bilge keel Lateral Hydrodynamic Loads on Bilge Keels”, 26th
added mass effects in the context of roll ONR Symposium on Naval Hydrodynamics,
Rome, Italy, September 17-22, 2006.
experiments where the roll axis passes through the

288
 Session 8 Basic Theory & Roll Damping – Intact and Damaged Ships

Miller, R.W., Bassler, C.C., Atsavapranee, P. and Sarpkaya, T. and O’Keefe, J.L. (1996), “Oscillating
Gorski, J.J., (2008), “Viscous Roll Predictions for Flow About Two and Three Dimensional Bilge
Naval Surface Ships Appended with Bilge Keels Keels”, Journal of Offshore Mechanics and Arctic
using RANS”, 27th Symposium on Naval Engineering, Vol. 118, February 1996.
Hydrodynamics, Seoul, Korea, 5-10 October 2008.

Newman, J.N. (1977), Marine Hydrodynamics, MIT

Press.

289