You are on page 1of 8

AUV DYNAMICS: MODELING AND

PARAMETER ESTIMATION USING


ANALYTICAL, SEMI-EMPIRICAL, AND CFD
METHODS

E. A. de Barros* 1 , A. Pascoal**, E. de Sa***

* Department of Mechatronics Engineering and Mechanical


Systems, University of Sao Paulo, Sao Paulo, SP. Brazil.
Email: eabarros@usp.br
** Institute for Systems and Robotics (ISR) and Dept.
Electrical Engineering and Computers, Instituto Superior
Tecnico, Lisbon, Portugal. Email: antonio@isr.ist.utl.pt
*** National Institute of Oceanography, Dona Paula, Goa,
India. Email: elgar@darya. nio. org

Abstract: The paper addresses the problem of autonomous underwater vehicle


(AUV) modeling and parameter estimation as a means to predict the expected
dynamic performance of underwater vehicles and thus provide solid guidelines
during their design phase. The use of analytical and semi-empirical (ASE) methods
to predict the hydrodynamic derivatives of a large class of AUVs with conventional,
streamlined bodies is discussed. An application is made to the estimation of the
hydrodynamic derivatives of the MAYA AUV, an autonomous vehicle that is being
developed under a joint Indian-Portuguese project. The estimates are used to
predict the behavior of the vehicle in the vertical plane and to assess the impact
of stern plane size on its expected performance. Copyright © 2004 IFAC.

Keywords: Autonomous Underwater Vehicles, AUV Modeling, Parameter


Estimation, Computational Fluid Dynamics

1. INTRODUCTION use of analytical and semi-empirical methods to


predict the hydrodynamic derivatives of a large
The paper addresses the problem of autonomous class of AUVs with conventional, streamlined bod-
underwater vehicle (AUV) modeling and parame- ies An application is made to the estimation of
ter estimation as a means to predict the expected a set of hydrodynamic derivatives for the MAYA
dynamic performance of underwater vehicles and AUV, an autonomous vehicle that is being de-
thus provide solid guidelines during their design veloped under a joint Indian-Portuguese project,
phase, well before they can be tested at sea. The see Figure 1. These estimates are used to predict
main core of the paper provides a roadmap to the the behavior of the vehicle in the vertical plane
and to assess the impact of stern plane size on its
expected performarrce.
1 Work supported in part by the Portuguese FCT POSl
Programme under framework QCA III and by project This work should be viewed as part of a long
MAYA of the Ad!. The work of the first author was term research effort that aims to contribute to the
supported in part by the Brazilian Ministry of Education
development of computational methods for com-
through a "CAPES" scholarship.

CAMS 2004 369


bined plant/ controller optimization (PCO), that
is, for the combined design of AUVs and respective
controllers to achieve increased performance at
sea while meeting stringent energy requirements.
See for example Silvestre et al. (1998), and the
references therein for an introduction to combined
PCO methods in the field of marine robotics. Cen- Fig. 1. A diagram of the MAYA AUV (NIO
tral to the development of efficient PCO design design); the fin arrangement is not shown
methods is the availability of procedures to esti-
mate the hydrodynamic parameters of an AUV tained in Section 2, Section 4 provides a study of
before it can be actually built and tested. the dynamics of the AUV in the vertical plane and
of the impact of the position and size of the stern
Methods for parameter estimation based on the
planes on open loop performance. Finally, Section
geometry and mass distribution of marine vehicles
5 provides a critical review of the results obtained
have been used for decades in the ship building
and discusses issues that warrant further research.
industry. Important steps have also been taken
in order to adapt parameter estimation methods,
originated in aeronautics , to the prediction of 2. ESTIMATION OF HYDRODYNAMIC
DERIVATIVES USING ANALYTICAL AND
submarine and AUV dynamics (Maeda and Tat-
suta, 1989; Bohlmann, 1990). Recently, spawned SEMI-EMPIRICAL(ASE) METHODS
by the widespread availability of powerful comput- Analytical and semi-empirical methods for the
ers , there has also been a surge of interest in apply- estimation of the hydrodynamic derivatives of
ing Computational Fluid Dynamics (CFD) meth- marine vehicles are well rooted in the theory of
ods to the prediction of stability derivatives for hydrodynamics. The use of a particular method
airplanes and marine vehicles (Humphreys, 2001). is decided by taking into consideration the phys-
However, to the best of our knowledge, no in- ical nature of each of the parameters to be esti-
depth , systematic studies have been done on the mated, together with the underlying simplifying
evaluation and validation of the above methods assumptions adopted when modeling the vehicle.
for AUV parameter estimation. As a consequence, In what follows, the computation of the hydro-
there seems to be lacking an established approach dynamic derivatives of a fully submerged body is
for AUV parameter estimation allowing for the organized by groups according to the nature of
computation of the modeling errors incurred. It the physical phenomena involved. The presenta-
is therefore important to try and compare the tion is restricted to those methods that were used
types of estimates that are obtained with ASE to predict the hydrodynamic derivatives of the
and CFD methods and to later judge the precision MAYA AUV. However, the methodology adopted
of those estimates by resorting to towing tank for parameter estimation applies to AUVs with
experiments. This type of information will cer- slender bare hulls that are solids of revolution and
tainly play a major role during the phase of vehicle with fin-type control surfaces at the stern.
design to meet desired open loop performance
requirements. At the same time, once bounds are In what follows we restrict ourselves to motions
known for the inaccuracies that are inherent to in the vertical plane and adopt the usual notation
prediction methods (and thus for AUV parame- for marine vehicles described in SNAME (1950).
ter uncertainties), better control methods can be Methods are described for the computation of the
devised to explicitly deal in closed-loop with the force and moment experienced by a fully sub-
uncertainties of the design models obtained. merged slender body when its motion is restricted
to small perturbations about steady motion in the
Motivated by the above considerations, and as vertical plane (that is , about the equilibrium con-
a contribution towards meeting the above goals, dition that corresponds to the vehicle moving for-
the paper guides the reader through the steps ward at a fixed speed U , with the angle of attack
involved in the estimation of the parameters of et, pitch rate q, and pitch angle () set to zero). The
slender body AUVs in the vertical plan, using end result is the set of so-called stability deriva-
information available from a number of sources, tives for the submerged body (SNAME, 1950).
mainly the Datcom stability and control hand- For the sake of clarity, some basic notation is
book (Hoak and Finck, 1978) . The paper is or- now introduced with the help of Figure 2, where
ganized as follows. Section 2 introduces the main a torpedo-shaped AUV is depicted. The vehicle
concepts and formulas that are at the root of consists of the main hull with length L and di-
the ASE parameter estimation methods used to ameter d, together with nose and tail sections,
compute a full set of hydrodynamic derivatives for and two stern planes that can be deflected by the
the MAYA A UV in the vertical plane. Section 3 same angle 8e . The vehicle's body frame with axis
discusses very briefly preliminary results of CFD {x, z } is centered at point O. It is convenient to
analysis of the vehicle. Using the parameters ob- define x* as the axial coordinate of a generic point

370 CAMS 2004


vehicle when it moves at a small angle of at-

L- Jt
tack. As it is customary, the computation of the
stability derivatives will be done by computing
the lift, drag, and moment terms that are natu-
rally expressed in the flow axis, and map those
8 into the body axis, in a non-dimensional form.
As a simple example, consider the expression
Z(a) = -[L(a)cos(a) + D(a)sin(a)] that relates
the body-axis force Z with lift L and drag D. Sim-
ple computations show that the hydrodynamic
z
i
derivative Z~ = D~~Q) , where q is dynamic pres-
Fig. 2. Coordinate System Adopted sure, equals -(CD(O) + CLJ, where CD(O) is the
drag coefficient for a = 0 and the lift coefficient
along the x- axis of the vehicle, as measured from CL", = DC;~Q) Q=O denotes the derivative of the
the nose tip, positive aft. The axial coordinate of normalized lift curve CL (a) at zero. Thus the
the nose base section is denoted xiv , while that of importance of computing lift as a function of angle
reference point 0 is called xo. We warn the reader of attack and drag at zero lift angle. Identical
that we use the terms fins to refer to the all moving considerations apply to the computation of the
planes at the stern of the AUV. However, the other hydrodynamic derivatives in Table 1. The
hydrodynamic paramet ers that account for the organization of the section reflects the different
effect of those surfaces take the subscript W (from steps involved in the computation of lift and mo-
wings) because this will simplify the consultation ment, as well as drag.
of related, relevant literature on aerodynamics.
2.2.1. Lift on the Bare Hull
2.1 Added-Mass Coefficients
The computation of the lift term for a slender
The added-mass coefficients allow for the com- body moving at an angle of attack with respect
putation of the forces and moments exerted on to the fluid requires careful consideration. In fact,
a body as if it were moving in an ideal fluid if one were to make the simplifying assumption
(Newman, 1977). A number of reliable methods that the motion took place in an ideal fluid, then
for their estimation are available in the literature. the theory of slender body hydrodynamics would
In their essence, all methods rely on the compu- show that a body with a pointy nose and tail
tation of a scalar velocity potential function from would produce a zero lift force . However, this
which the velocity of the particles (that is, the does not occur in practice because viscous effects
flow) around a marine vehicle can be derived. In induce the appearance of vortices and changes in
the case of simple bodies, the flow can be obtained the pressure distribution at the stern, even in the
by combining the velocity potentials due to a case of small angles of attack. As a result, lift
distribution of sources and sinks. These results occurs in a non-ideal fluid. Experimentally, it is
have been tabulated for a number of conventional verified that the pressure distribution at the fore
shapes, assuming the motion takes place in an part of a streamlined body agrees quite well with
unbounded fluid. For other types of slender ve- the ideal flow prediction(Hoerner, 1985) . However,
hicle shapes, the three-dimensional added-mass the pressure at the after body is reduced over
coefficients can be approximated by a strip-theory a region that starts at a point where the vortex
synthesis, that is, by integrating bi-dimensional production or the boundary layer separation takes
added-mass coefficients for each section along the place and progresses all the way to the rear end of
vehicle's length. the body. The transition point at the after body
is likely to occur at places where the change in
In the case of the MAYA AUV, the added mass the hull slope is bigger. Taking this phenomenon
coefficient in surge was computed by fitting ap- into account , the formula proposed by the US Air
proximately an ellipsoidal shape to the vehi-
Force Datcom (Hoak and Finck, 1978) to compute
cle's hull. The other added-mass coefficients for
lift on a slender body considers the same value of
the body and fin ensemble were computed using
the lift coefficient as that obtained in the ideal flow
strip-theory and the formulas for bi-dimensional case, but taking only in consideration the length
added-mass coefficients of circles and finned cir- of the body from the tip of the nose to the point
cles (Newman, 1977) . The results are tabulated
where the ideal flow hypothesis is no longer valid.
in Section 4.
In the case of missiles or torpedo shapes AUVs like
2.2 Static Coefficients MAYA, the region where ideal flow predictions are
accurate is restricted to the nose. In this case, the
This section describes the computation of the bare hull lift coefficient CLOB (normalized by the
force and moment exerted by the fluid on a square of the vehicle length L), is given by

CAMS 2004 371


that builds on the methodology derived in Pitts
C = (8CLB) =2(k2_k1)SN (1)
LOB 8a L2 et al. (1957) for aircraft and missiles. The key
<>=0
idea is to compute the so-called interference fac-
where C LB denotes lift as a function of a and SN
tors between the lift surfaces and the body using
is the cross section area at the nose end coordinate
slender-body theory. The first step in this pro-
x'N. In the above expression, k2 - k1 is the "Munk" cedure assumes that the fins do not deflect, the
apparent mass factor , which is a function of the
effect of the deflection being taken into account
fitness ratio f = ~ of the body. For values of the
at a later stage. This step yields an approximate
fitness ratio 4 ~ f ~ 19, the "Munk" factor can
expression for the total (body plus fins) lift coef-
be approximated by the polynomial interpolation
ficient CL a (WB ) (where the notation W B borrows
(k2 - k 1) = -0.0006548 f2 + 0.0256 f + 0.73. (2) from aircraft wing-body interactions), given by
Under the same simplifying assumptions, the mo-
CL
a(WB)
= CL
OB
+ CL a W ( B) + CL a w) (5)
ment coefficient (normalized by the cubic of the
S
vehicle length) is given by: = C Lo B + (KW(B) + KB (W »)(CL,, )e L~
C
mUB
- 2(k2 - k1)
- L3
J
x;.,.
dS x '* * *
dx* (x - xo)dx , (3)
where Se is the total exposed fin surface area,
(CL a )e is the lift coefficient of the exposed fin
o surfaces, and KB (W ) and KW(B ) are the interfer-
where Sx' denotes the body section area at x*. ence factors from the surfaces to the body, and
From the above, it follows that the hydrodynamic from the body to the surfaces, respectively. Let b
center of the body, normalized by L and expressed be the maximum span of the fins in combination
in body-axis coordinates, is located at x' with the hull, that is, the total distance between
Cm" B B control surface tips as if they extended inside the
CL U B ' hull, and define
2.2.2. Lift Produced by Small Aspect Ratio Fins d
k= t; . (6)
Then, the interference factors can be written as
Studies on the lift force CL produced by small 2
K _ 2 (1 - k4) (1 - k (2
aspect ratio fins have been conducted by many WeB ) - (7)
researchers and have led to closed formula approx- - 11" (1 - k)2
imations that are widely available in the litera- and
ture. In this study we adopt the approximation
for CLow = (a~~w) proposed by Whicker
KB(w) = (1 + k)2 - KW(B), (8)
<>=0
where
and Fehlner (1958) and used by Bohlmann (1990)
to compute lift in submarine hydroplanes, given (1 = [~tan-1(~(k-1 - k)) +~] (9)
by
and
C Low 21f
(4) (2 = [(k- 1 - k) + 2tan- 1 k].(10)
AR 2+ v#d co j/;\:/4 + 4cos 2
A c/ 4 ) The estimation of the corresponding total moment
coefficient Cm O( W B) is given by the product of
where AR is the lift surface aspect ratio, Ac/4
is the sweep angle at one fourth of the chord the lift coefficient computed above and the po-
length, and TJ is a factor to correct for viscous sition X(WB) of the new hydrodynamic center,
effects (adopted as 0.9). The above expression normalized by L, computed by taking into con-
yields results that are very close to the to the ones sideration the interaction between the hull and
described in Hoerner (1985) for small aspect ratio the fins. To compute x(WB)' start by defining
surfaces. Classical results can be used to compute X:V (B ) and x~(W) as the cent er of the hull-lift
the hydrodynamic center of the fin (Abbot and carryover on the lift surface and the center of the
Doenhoff, 1949) . In what follows , x:V denotes the fin-lift carryover on the body, respectively. The
axial coordinate of the hydrodynamic cent er of the first can be taken as approximately equal to the
AUV fins, in body-axis. hydrodynamic center of the fins x~. The latter is
given by
, I b- d
2.2.3. Fins and Body Combination xB(W) = 4 + -2- tan Ac/ 4 * P, (11)
cr e
There is a mutual influence between the fins and
where
the main body of an AUV due to the changes
k
in the flow past each of these components caused P=-l-k+
by the presence of the other. To account for the
impact of these effects on the total lift experienced
by a fully submerged body, we follow the proce-
J1-2kln( ¥ + iJ1-2k) - (1- k) + ~k
dure described in Datcom (Hoak and Finck, 1978) k ( l k)
v'1 - 2k
In( 1-k k + 1.k VI
/f=2k)
- '('r.;
+ (1 - kk ) 2 _ 2!:(1
2
- k)

372 CAMS 2004


where Cr. is the exposed tip root chord, and -(CDo + CL"' (WB»
Ac/4 is the sweep angle at one fourth of the root -(CL q +X~)
chord length. This expression above assumes that Cm"' (WB )
the aspect ratio is greater than or equal to 4. C mq
For smaller values of the aspect ratio, the reader - CL 6e
should consult an interpolation procedure used at C m6•
Datcom. It is now possible to compute Table 1. Equivalence between stabil-
C L "' W (B ) x'W (B) +CL "' B(W ) x'B(W ) +Cm "' B ity derivatives and hydrodynamic coef-
I
X(WB) = C ficients
L"'(WB )
(12) mass (m) 47.5 (Kg)
To capture the effects due to the deflection of moment of inertia (Iyy) 8.923 (Kgm 2 )
reference length (L) 1.64m
the control surfaces, the methodology described main diameter (d) 0.2m
in Datcom leads to the control surfaces lift and tail span (b) 0.44m
torque coefficients C L6e and C m6 e , respectively tail exposed area (Se) 0.018(m2)
given by tail and hull combined area (Sw) 0.0385(m2)
tail aspect ratio (AR) 5.02
Se tail exposed aspect ratio (ARe) 3.2
C L6 e = (kB(w) + kW(B ) )(CL", )e L2 (13)
tail exposed taper ratio (A) 0.667
and cruising Speed (U) 1.5(m/s)
C m 6e = CL 6e X~ (14) Table 2. Particulars of Maya
where be is the deflection angle of the lift surfaces
and x~ is the center location of the exposed lift m' 21.145 M' -0.612
surfaces. It is important to remark that kB( w ) +
Q

I~y 1.48 M'q -1.677


kW(B) = KW(fJ) X'u -0.478 M' -2.323
Z'q -0.612 M'"q -3.221
Zl. -24.138 M'8e -8.089
2.2.4. Drag Coefficient Z'" -68.862 Z'8e -30.615
Z~" -11.712 M~ -1.84
This section details the computation of drag coef-
ficient of an AUV at zero angle of attack. As dis- Table 3. The Maya Derivatives (*10- 3 )
cussed before, this is the only drag-related infor-
mation required to compute the stability deriva- moment, respectively to variations in the pitch
tive Z~. In this case, drag equals the pressure plus angular velocity. Simple approximations to the
friction drag. The bare hull drag coefficient is cal- coefficients can be obtained by emphasizing the
culated in Datcom as a function of the Reynolds contribution of the control surfaces, yielding
number Re, the fitness ratio 1, and the base diam-
eter db. To compute the zero-lift drag coefficient C Lq = -(KB(w ) + KW( B»)(CLQ ) e ~~X;¥( B) (19)
C D o , start by defining

CD = C f [l + 60r 3 + 0.00251] ~~ (15)

where Ss is the total body wetted area and Cm. = - (KB (W ) + KW( B»)(CLQ ) e ~~(X:"( BY
0.075 (20)
C f = (log Re _ 2)2 + 0.00025 (16)

is the skin friction drag coefficient, as given by the


ITTC. Further define the base-drag coefficient 2.4 Stability Derivatives

CD b = 0.029(~)3(CD)-O.5 . (17) As explained before it is customary, when mod-


eling marine vehicles, to parameterize the mod-
where db is the base diameter, that is, the diame- els in terms of so-called stability derivatives, as
ter of the sternmost section of the body. Then, described in SNAME. The equivalence between
CDo =C D+ C D b (18) non-dimensional stability derivatives and the hy-
drodynamic coefficients determined before is sum-
Results on the fin drag coefficient are provided by marized in Table 1 (see also Blakelock (1991) for
Abbot and Doenhoff (1949). a lucid exposition of the subject in the aircraft
area) . In preparation for the study that follows,
the particulars and stability derivatives of the
2.3 Dynamic Coefficients
MAYA AUV in the vertical plane are given in
The dynamic coefficients C Lq and Cmq of a sub- Tables Table 2 and Table 3, respectively. The bare
merged body relate the heave force and pitch hull shape is based on the geometry proposed in
(Myring, 1976).

CAMS 2004 373


3. CFD BASED METHODS (m' - Z~)L/U ao (I'yy - M:)(L/U)'l.
z' bo (M~ - m'x'c;)L/U
To estimate the hydrodynamic coefficients of (Z'q + m'x"~)(L/U)2 co (M~ + m'x~)L/U
(Z~ + m')L/U do M'
MAYA, Computational Fluid Dynamics (CFD)
eo Mo"
methods are also being exploited by the National
Institute of Oceanography in Goa, in India which Table 4. Transfer function coefficients
has access to the FLOWSOLVER machine at
Bangalore. At the time of writing of this paper, the center of mass. From the above equations, it
only results on the drag coefficient at zero-lift follows that
angle were available. Depending on the convec- o:(s) N",(s)
(24)
tion and turbulence scheme used, the drag coef- oe(s) D(s) ,
ficient(based on the maximum section area) was B(s) Ne(s)
found to vary between 0.08 and 0.127. The same (25)
oe(s) D(s) ,
coefficient calculated using Datcom yield 0.143. A
where
careful analysis of these results and the extension
of CFD methods to compute the lift related coeffi- , [(1-xw-
N", (s ) = aeZlie , C"') s 2 +(be , d",) s- -ee 1,
---Xw-
cients are planned for future work. Together with ae ae ae ae
(26)
real data from tank or open ocean tests, this study
is expected to shed light into the accuracy of the ce )s + (b",
Ne (s ) = Mli, ea", [(1 - -,-- -- de
- --,-)],
results that can be obtained with CFD methods
xwa", a", a",x w
(27)
for AUVs of this type.

4. DYNAMIC ANALYSIS
The set of stability derivatives obtained above is (28)
now used to predict the dynamic behaviour of The denominator was simplified by neglecting the
the AUV in the vertical plane and to evaluate products Cec"" cedo:, and co:de when compared to
the impact of stern plane size on its expected the other terms contributing to the coefficients of
performance. the same power of s. This is a consequence of the
body symmetry characteristics, which imply small
4.1 Open loop transfer functions values for the added mass terms Z4
and M~. The
term D(s) can be further simplified by noticing
To study the dynamics of the AUV in open that Ido: de I « I-eea", + b",bel. This follows from
loop, it is sufficient to compute and analyze the the following observations: for vehicles of the
transfer functions from stern plane deflection to type considered d", is generally positive, with
pitch and heave motions. Let o:(s), O(s), and oe(s) magnitude less than half of a",. Furthermore,
denote the Laplace transforms of 0:, 0, and oe, de (which equals the static moment coefficient
Cm ~ ) has the same order of magnitude or less
respectively. Neglecting the surge equation, the " ( W B)

linearized model of the AUV in the vertical plane when compared to ee and be. Finally, b", has a
(about its steady forward motion at trimming large magnitude when compared to the previous
speed) is easily seen to be given by coefficients. With the simplifications above, D(s)
can be written as
(a", s - b",) o:(s) + (-c",s2 - d",s)B(s) = Z~eoe(s)
(21) D(s) ~ a",ae(s - b",) [s2 _ be s + ~( - ee)] .
a", ae ae
and
(29)
(- ces - de) 0:(s) + (aes2-bes - ee)B(s) = M~eoe(s), The first order term in D(s) captures the heave
(22) dynamics, while the second order term is related
where the relationship between the hydrodynamic to the pitch dynamics.
derivatives and the coefficients above are given The small Cm U( W B) hypothesis allows for a further
in Table 4. Further let ib (s) denote the Laplace
simplification in the pitch transfer function. To see
transform of depth rate in non-dimensional form,
this, start by noticing that the pitch motion can be
where depth is measured from the surface, posi-
practically de coupled from the heave dynamics by
tive downwards. Linearizing the depth coordinate
neglecting the coefficient Ce and the term ~ in
dynamics about trimming yields aaX UJ

Ne (s). The last simplification becomes clear when


ib(s) = o:(s) - O(s) (23) the last term in Ne(s) is re-written as
In order to simplify the analysis, assume that (_ b", _ ~) = _ bo: (1 +, X'(~B)), (30)
the origin 0 of the body axis is coincident with ao: ao:xw a", Xw

374 CAMS 2004


where 0 < 'Y < 1 due to the drag coefficient con- pitch acceleration at time zero. To compute this
tribution in Z~, see Table 1. As long as ICm a(WB) I ratio, use the transfer functions defined before and
is small enough, so is the ratio between the hy- apply the initial and final value theorems to the
drodynamic center of the vehicle and that of the calculated step responses of pitch and heave to
fins, and the expression above degenerates approx- obtain
imately to - ~. As a consequence, there is an q(O) = M~e~a = -x:V ba (33)
almost pole-zero cancellation (in the pitch transfer ass aoZoe ao
function) tied with the heave dynamics. Although In order to simplify the analysis, suppose that the
the present version of Maya does not have a very problem is limited to searching the optimal span
small value of ICma(WB) I, the simplifications above and location of the stern planes, taken as two rect-
result in reasonable approximations to the trans- angular, non-cambered fins . Further assume that
fer functions . In fact, the poles obtained from the modification of the fins does not change the
expression (28) are center of mass of the vehicle significantly. Notice
-1.166, - 0.571 ± 0.500i rads- 1 , that the changing in the span of the fins affects
both ao and ba . When the span is increased, the
while those obtained from the simplified expres- change in ba is mainly due to an increase in the
sion (29) are magnitude of the lift force, whereas the change
in ao is mainly due to an increase in magnitude
-1.373, - 0.467 ± 0.637i rads- 1
of the added-mass coefficient M 4. To compute the
The zero in the pitch transfer function is -1.198 change in the latter coefficient, start by computing
rads- 1 . From (23)-(30) and the expressions in the sectional (heave) added mass coefficient of a
Table 4, it is also possible to derive the transfer finned circle, given by (Newman, 1977)
function from stern plane deflection to depth rate
as (34)
ib No(s) (31)
c5e D(s) ' where be = b - d is the total exposed fin span.
where It is then straightforward, using strip theory, to
() =ao
N DS ~ Z' [2 LZ:'x\,y
oes - -
o
- - s - x\,yba+e ]
U ao ao
obtain the contribution of the fins to the new value
of M 4. This is done by multiplying the sectional
(32) added mass coefficient above by the square of the
distance from the section to the origin 0 of the
By examining the signs and relative magnitude of
body-axis and integrating the contributions over
the coefficients involved, it follows that the depth
the lift surface chord. Clearly, from (34), for a
dynamics exhibit non-minimum phase character-
fixed position of the hydrodynamic center of the
istics.
fins, the variation in M4 (and therefore in ao)
is practically proportional to the square of the
4.2 Maneuvering performance surface span. The change in ba takes place in a
more complex form, as the discussions in sections
Since the hydrodynamic derivatives depend on the 2.2. 2 and 2.2.3 show.
overall hull geometry and fin arrangement of an
The performance index must also take into ac-
AUV, they are expected to play a key role in the
count the increase in energy consumption of the
development of methodologies for vehicle design.
vehicle in steady motion (that is, with zero angle
In this paper, and as an illustrative example, we
of attack) due to an increase in fin span. Clearly,
consider the problem of optimizing the maneu-
this will be due to added drag. Since the fins are
verability of an AUV in the vertical plane, when
rectangular and have a fixed chord, the added drag
the bare-hull profile is fixed , by proper choice of
force (at zero angle of attack) is proportional to
the stern plane dimensions and location. Clearly,
the fin span. As a consequence, the energy penalty
this entails the definition of a performance index
factor is inversely proportional to the fin span.
to capture the maneuverability requirements in a
In this study, the combined span of the fins was
rigorous manner, as explained below.
allowed to vary between 0.3m and 1.0m. In the
Suppose the objective it is to achieve a fast re- current design of MAYA, the corresponding figure
sponse in a surfacing/ diving emergency maneuver is 0.44m. With the largest span considered, the
(e.g. for collision avoidance), while keeping the stern plane contribution to the vehicle total drag
angle of attack small so as to stay in the region of rises to about 30 percent. It is also important
validity of the linear design model. In this case, that the performance index should penalize the
a possible choice for the performance index is difficulties in vehicle handling that arise when the
the ratio q'(O) /a ss , where ass is the steady state span of the surfaces assumes large values. Again,
value of the angle of attack in response to a stern this calls for a penalty factor that is inversely
plane step deflection and q'(O) is the resulting proportional to the span size. Based on the above

CAMS 2004 375


considerations, the final expression adopted for
the performance index is
1= 1 4'(0) (35)
(0.5 + 2b e ) ass
-6

where the factor that multiplies the ratio q'(O) is j


Q ss
~
• •
normalIzed III such a way as to yield the value 1 for 5

the current MAYA surface span, and slightly less


than 0.5 for the largest span. Figure 3 is a series of
i
0. 4

plots of the performance index versus total span,


for three different locations of the stern plane hy-
drodynamic center. The middle plot corresponds
to the original location (0.459m aft the center of 0.5 0.6 0.7 0.8 0.9
Total Span: b(m)
mass), whereas the two other plots correspond to
locations closer to the center of mass (-0.204m)
Fig. 3. Effect of stern plane span on maneuvering
and further away from it (-0.604m ). The latter
performance
case yields the largest performance index. Inter-
estingly enough, increasing the distance from the Hoak, D. and R. Finck(1978). USAF Stability and
fins hydrodynamic center to the cent er of mass of Control Datcom. Wright-Patterson Air Force
the vehicle, allows for smaller optimal fin sizes. Base, Ohio.
There is no advantage in increasing the size of the Hoerner, S. (1985). Fluid Dynamic Lift. Hoerner,
fins past the optimal value of a particular curve, Liselotte A.
because the added moment of inertia takes over Humphreys , D. (2001). Correlation and Valida-
and forces the performance index curve to slope tion of a CFD Based Hydrodynamic & Dy-
down. namic Model for a Towed Underwater Vehi-
cle. Proc. MTS/ IEEE OCEANS 2001, Hon-
5. CONCLUSIONS AND olulu, Hawaii.
RECOMMENDATIONS FOR FUTURE Maeda, H. and S. Tatsuta (1989). Prediction
RESEARCH. method of hydrodynamic stability derivatives
of an autonomous non-tethered submerged
The use of analytical and semi-empirical estimates vehicle. Proc. of 8th Int. Conf. on Offshore
(ASE) for the hydrodynamic derivatives of AUVs Mechanics and Artic Engineering.
holds great potential to the development of power- Myring, D F (1976) . A theoretical study of body
ful tools for optimal vehicle design. The estimates drag in sub critical axisymmetric flow. Aero-
can be used to predict the dynamic behavior of the nautical Quarterly 27(3) , pp. 186- 194.
vehicle and to assess the impact of fin arrangement Newman, J N (1977). Marine Hydrodynamics. 9th
on its expected performance. It is therefore impor- ed., M.LT., Cambridge, Massachusetts.
tant to try and compare the types of estimates Pitts, William C, Jack N Nielsen and George E
that are obtained with ASE and CFD methods Kaattari (1957). Lift and cent er of pressure
and to later judge the precision of these estimates of wing-body-tail combinations at subsonic,
by resorting to towing tank experiments. transonic, and supersonic speeds. Technical
report, NACA.
ACKNOWLEDGMENT SNAME, The Society of Naval Architects and
The authors would like to thank A. Alcocer, H. Marine Engineers (1950). Nomenclature for
B0, and S. Fekri for their help with the text format Treating the Motion of a Submerged Body
and figures. Through a Fluid. Technical and Research Bul-
letin, No. 1-5.
Silvestre, P., A Pascoal, I Kaminer, and A Healey
REFERENCES
(1998). Combined plant/ controller optimiza-
Abbot, 1. and A. Von Doenhoff(1949). Theory of tion with application to autonomous un-
Wing Sections. Dover Publications Inc., New derwater vehicles. Proc. CAMS'98, Fukuoka,
York. Japan.
Blakelock, J. (1991). Automatic Control of Air- Todd, F. (1967). Resistance and Propulsion. Prin-
craft and Missiles. Second Edition, John Wi- ciples of Naval Architecture, Chapter VII.
ley Sons, Inc. John P. Comstock, Ed. New York.
Bohlmann, H. (1990). Berechnung Hydrody- Whicker, L F and L F Fehlner (1958). Free-stream
namischer Koeffizienten von Ubooten zur characteristics of a family of low-aspect-ratio,
Vohrhersage des Bewegungsverhaltens. PhD all-movable control surfaces for application
thesis. Institut fur Schifbau der Universitat to ship design. Technical report 933, David
Hamburg. Taylor Model Basin.

376 CAMS 2004

You might also like