Luca Formaggia

MOTIVATIONS

MOX, Department of Mathematics F. Brioschi

Politecnico di Milano, Italy
GEOMETRIC
PREPROCSSING

The study of the dynamics of a racing boat may help the

designer (shape optimization) as well as the trainers (crew
optimization)

GEOMETRIC
PREPROCSSING

IMA Workshop on Computing in Image

Processing, Computer Graphics,
Virtual Surgery, and Sports
Minneapolis, 7-11 March 2011

Trimming
Sea keeping

http://mox.polimi.it
MODEL
VALIDATION

MODEL
VALIDATION

Numerical modelling of competition rowing

boats
OUTCOME

Manouvering
Performance

OUTCOME

Rower positioning
Rowing style

Other contributors : Edie Miglio, Nicola Parolini, Andrea Mola, Anna Scotti, Andrea
Paradiso, Lorenzo Tamellini
IMA 2011

IMA 2011

An example of complex dynamics: accelerations

induced by the rowers action and movement

Comparing different styles (horizontal accelerations)

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

Excellent rower

Intermediate Rower

Images taken from On board shell measurements of acceleration by K. Young and R. Muirhead,
http://www.phys.washington.edu/users/jeff/courses/ken_young_webs/rowsci

OUTCOME

OUTCOME

Accurate simulation of the wave resistance caused by secondary motion

may contribute to the optimization of the boat and of the athlete action.
IMA 2011

IMA 2011

A multilevel approach
Boat
material
and
geometry

Rowers
motion and
weight

Dynamical system of rowing boat motion

MODEL
VALIDATION

Hydrodynamic forces

RANS
OUTCOME

GEOMETRIC
PREPROCSSING

Oarlock
forces

(VOF)

Free
surface
Hydrod.

3D
Potential
Eqs.

Strip
Theory

Hull Position

BOAT MOTION

GEOMETRIC
PREPROCSSING

An example obtained with a RANS code

MODEL
VALIDATION

Costly computations! We wanted a simpler models for

preliminary studies

OUTCOME

FREE SURFACE DYNAMICS MODEL

IMA 2011

IMA 2011

OUTLINE OF THE TALK

A typical rowing boat
GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Modeling the dynamics of the boat and the rowers

A potential model for fast computations
A model based on free surface dynamics and
unilateral constraint

MODEL
VALIDATION

MODEL
VALIDATION

Conclusions and further work

OUTCOME

OUTCOME

IMA 2011

IMA 2011

The dynamics of a rowing boat

Full 3D dynamics

Movements in the (x,y) plane

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Horizontal acc.
Gh

Boat reference system

MODEL
VALIDATION

OUTCOME

MODEL
VALIDATION

Pitch
sink

These are the most important movements. A first model

was developed accounting only for them

cos cos
R = cos cos
sin

OUTCOME

Absolute reference system

sin sin cos cos sin

sin sin sin + cos cos
sin cos

pitch angle

yaw angle
roll angle

cos sin cos + sin sin

cos sin sin sin cos
cos cos

X = Gh + Rx

IMA 2011

IMA 2011

Computation of the forces at footboards and seats

Modelling the oar
GEOMETRIC
PREPROCSSING

The oar has been modeled

as a simple lever

Y
X

-Fh
rh
-Fo

boat hull
water
L

MODEL
VALIDATION

A dynamical model for the rower

GEOMETRIC
PREPROCSSING

h = hand

j
This model together with the model for the oars allows
us to
write the forces at the seat Fs and at the footboards Ff as
function of the force at the oarlocks Fo and the rower
motion.

MODEL
VALIDATION

rw
Fw
OUTCOME

Fh =

L rh
Fo
L

OUTCOME

p=12
j=1,..n

IMA 2011

IMA 2011

The reconstructed kinematics

The kinematics of rowing

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

The kinematics of the rowing is extracted

from a data base of measurements made
using video capturing techniques

OUTCOME

OUTCOME

In collaboration with C. Sforza of the Istituto di morfologia umana of Universit di

Milano, Italy
IMA 2011

IMA 2011

The dynamical model of the boat-rowers system

Forces at the oarlocks

GEOMETRIC
PREPROCSSING

Oarlock forces may be reconstructed from experimental

data. We need to separate the active phase from the
recovery phase of the rowing action.

Fo,x =
MODEL
VALIDATION

f1 (t) 0 t < a
f2 (t) a t < T

GEOMETRIC
PREPROCSSING

Footboard

+ I1 (, ;
xr ) =
MG

Fo,z = Fo,x

2n

j=1

Oarlock

Seat

F oj + M g + F w

MODEL 1
RI
+ RIG R1 + I2 (, ;
xr ) =
GR
VALIDATION

Lrh
L

The vertical component is taken

proportional to the horizontal
OUTCOME
one, while f1 and f2 are
approximated by a cubic and
quadratic polynomial in t,
respectively

o oarlock


2n 
h
F oj +
j=1 X olj G


2n 
h
F oj + M w
j=1 X hlj G

rotation vector Gh boat center of mass IG tensor of inertia

L is the oar length and rh the part from the oarlock to the hand

OUTCOME

Fw and Mw are forces and momenta due to the fluid.

IMA 2011

IMA 2011

The main hypothesis

Reduced model for the hydrodynamic interaction

Computations with RANS codes are expensive in terms of

human and computer resources

GEOMETRIC
PREPROCSSING

In the preliminary design phase there is the need of a fast, yet

effective tools to compare different configurations
A fast tool can be used also by athletes and trainers to test
different rowing styles or boat arrangements

MODEL
VALIDATION

Main viscous phenomena are captured by static simulations at

constant speed

GEOMETRIC
PREPROCSSING

Secondary motions dissipate energy mainly through wave

radiation
The length of the waves generated by the secondary motions
is comparable to that of the boat while the amplitude is small
compared to the wave length

MODEL
VALIDATION

We can neglect non-linear phenomena when computing the

effects of the secondary motions

OUTCOME

OUTCOME

A first reduced model: we solve just ODEs (coupled with an

off line Laplace equation)!

Secondary motions are periodic with period T=2/ equal to

the stroke period

IMA 2011

IMA 2011

Decomposition of the hydraulic forces acting on the hull

Simulation of secondary motion effects

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

The secondary motion

of the boat is considered as formed
by elementary periodic movements

Drag force

is estimated from off-line static computations

MODEL
VALIDATION

The induced velocity field is described by a potential in the

complex plane
MODEL
VALIDATION

Bouyancy force

OUTCOME

By linearising the free surface interface conditions and

applying first order radiation condition at the artificial bounary
we obtain a set of Laplace equations in the complex plane

OUTCOME

is the wet surface, dynamically computed

IMA 2011

IMA 2011

The potential problem

The forces induced by the secondary motion

GEOMETRIC
PREPROCSSING

h , )
v = (G

GEOMETRIC
PREPROCSSING

Added mass matrix

Damping matrix
To account for more frequencies we need to solve a convolution
integral

MODEL
VALIDATION

MODEL
VALIDATION

F D = M()v +

is a the generalised normal

(in the 3 d.o.f. case)
OUTCOME

OUTCOME

K(t) =

1
Re

)d
K( )v(t
0

(M() M()) eit d

Pitching movement

C.C. Mei, The applied dynamics of ocean surface waves, Wold Scientific, 1989
IMA 2011

Vertical movement
IMA 2011

The final system of equations

The control on sway and roll

The system is unstable in the roll degree of freedom and indifferently

stable on the yaw degree of freedom. We have added a simple
feedback control to simulate the action of the rowers

GEOMETRIC
PREPROCSSING

Fo,z


0 0 < t a
=
kroll a < t T

MODEL
VALIDATION

GEOMETRIC
PREPROCSSING

Eventually, we have a system of non linear second order

ODEs for the linear and angular displacements u

= f (u(t), t) t > 0
M(u(t), t)
u(t) + K(u(t), t)u(t)
with given initial position and velocities. We reduce it to a first
order system


u
y
=
M (t, y)y = F (y, t).
u

MODEL
VALIDATION

Fo,x =

kY aw 0 < t a
0 a < t T

solved with a standard RK45 scheme

OUTCOME

OUTCOME

IMA 2011

IMA 2011

Example: single scull 80 Kg athlete

Comparison: heavier vs ligher rower

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

OUTCOME

Surge

Heave

Pitch

Speed

Heave

Poisition

GEOMETRIC
PREPROCSSING

IMA 2011

IMA 2011

Sensitivity Study

The importance of the control

GEOMETRIC
PREPROCSSING

Polynomial chaos expansion

P

(x, t, )
i (x, t)i ()

GEOMETRIC
PREPROCSSING

Oarlock Force

i=0

FXmax
Rowing cadence

MODEL
VALIDATION

1
T

Multi-physics Model

MODEL
VALIDATION

VX

T

G hX (t)dt

Efficiency

Rower mass

OUTCOME

Mean surge velocity

OUTCOME

mr
Input PDF
IMA 2011

Output PDF

IMA 2011

An example of sensitivity analysis

The code in action

Male coxless four. Sensitivity study

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Parameters: Horizontal oar force (4), rower weight (4), cadence

MODEL
VALIDATION

OUTCOME

Input Param.
FXmax 1
FXmax 2
FXmax 3
FXmax 4
mr 1
mr 2
mr 3
mr 4
r

VX sensit.
8.38 104 ms /N
8.41 104 ms /N
8.34 104 ms /N
8.37 104 ms /N
0.0066 ms /Kg
0.0064 ms /Kg
0.0065 ms /Kg
0.0063 ms /Kg
0.0343

m
/( strokes
)
s
min

sensit.
0.0014 1/N
0.002 1/N
0.0015 1/N
0.0003 1/N
0.0287 1/Kg
0.0176 1/Kg
0.0142 1/Kg
0.0086 1/Kg

The model is currently used at Filippi Lido srl, a renown

rowing boat manufacturer, for preliminary design and boat
trimming.
www.filippiboats.it

MODEL
VALIDATION

OUTCOME

0.1589 1/( strokes

)
min

Sensitivities at point = 0
IMA 2011

IMA 2011

The problem setting

An alternative hydrodynamic model

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

At an intermediate level of complexity between full RANS

simulations and the potential model just described we have
hydrodynamic free surface models which describes the
surface elevation explicitly.

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

OUTCOME

IMA 2011

IMA 2011

Kinematical conditions

DAlambert-Lagrange principle

The flow motion satisfies at any time t the following relation

The velocity is indicated as

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Incompressibility
where P is any admissible virtual
particle displacement

Free surface evolution

Impermeable wall
MODEL
VALIDATION

Non slip condition

at the bottom

Other conditions will be

examined later on

MODEL
VALIDATION

The corresponding differential equations are

Momentum equation

OUTCOME

OUTCOME

Boundary
compatibility
cond.
Natural boundary
conditions

`
a atmospheric pressure, W external surface forcing term
IMA 2011

IMA 2011

Imposing of the presence of a boat by a inequality

constraint

Lagrange multiplier technique

We introduce a Lagrange multiplier

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

satisfying

and we add to the rhs of the variational formulation the term

MODEL
VALIDATION

Constraint

MODEL
VALIDATION

which represents the virtual work done by leading to

OUTCOME

OUTCOME

IMA 2011

IMA 2011

The resulting surface Navier Stokes equations with

unilateral constraints

Intermpreting the augmented variational formulation

The result is that the boundary condition on s becomes

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Note that on s
by which

MODEL
VALIDATION

MODEL
VALIDATION

It is convenient to rewrite the pressure term as

hydrostatic pressure
OUTCOME

OUTCOME

hydrodynamic correction

constraint reaction term

IMA 2011

IMA 2011

An important note

Specialising the model

The kinematic relation for together with the definition of N on s

implies that on we have the following slip condition on the
velocity

GEOMETRIC
PREPROCSSING

that is no conditions are imposed on the tangential component

Indeed fixing the tangential component of the velocity on the
boat surface is would be incompatible with the kinematics
of the free surface and the condition =
on

MODEL
VALIDATION

Therefore the effect of friction on the boat may be only take

into account empirically through the function W

OUTCOME

IMA 2011

Under the hypothesis that the basin is relatively shallow we

can make the following approximation

GEOMETRIC
PREPROCSSING

and neglect all horizontal components of the stress. This

approximation leads to very convenient numerical schemes
(but it is not crucial for the numerical algorithm for the
imposition of the constraint)

MODEL
VALIDATION

Furthermore, by integrating along the vertical the kinematic

condition for and using the continuity equation we have
OUTCOME

IMA 2011

10

The final set of equations

Numerical treatment of the time derivative

At each time step

GEOMETRIC
PREPROCSSING

we set

GEOMETRIC
PREPROCSSING

Flow equations

where

is obtained by solving

Free surface
kinematics
MODEL
VALIDATION

MODEL
VALIDATION

Boundary
Conditions

with a suitable time integration scheme

OUTCOME

OUTCOME

CD is the Chezy coefficient. It accounts for friction at the bottom surface.

We avoid resolving the boundary layer.
IMA 2011

IMA 2011

Time discrete equations

Weak formulation setting

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

We define the following forms

MODEL
VALIDATION

MODEL
VALIDATION

Linearization procedure
OUTCOME

OUTCOME

IMA 2011

IMA 2011

11

The b term explained

The weak formulation

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

OUTCOME

(a,b) denotes the L2(

) scalar product
IMA 2011

IMA 2011

Interpratation of the weak formulation

Numerical solution of the constrained problem

The weak formulation may be reinterpreted as the KKT condition

for the following minmax problem

GEOMETRIC
PREPROCSSING

where

By the strong duality principle the problem is equivalently

stated as
where

GEOMETRIC
PREPROCSSING

which can be implemented as a succession of

unconstrained problems.

MODEL
VALIDATION

MODEL
VALIDATION

associated to the following minimization problem

In particular a projected steepest-descent type method

applied to the problem in w leads to the well known Uzawa
iterations

OUTCOME

OUTCOME

J. Cea, Lectures on Optimization Theory and Algorithms, Tata Institute, 1978

IMA 2011

IMA 2011

12

Finite element space discretization

Uzawa iterations

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

This choice
allows to get an
efficient and
easy to
parallelize
scheme

OUTCOME

E. Miglio, A. Quarteroni and F. Saleri., CMAME, 1999

IMA 2011

IMA 2011

The algebraic setting

A fractional step scheme

Hydrostatic iterations:

For every iteration of the Uzawa scheme we need to

solve a system of the form

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Vertical velocity d.o.f.

Horizontal
Elevation
velocityd.o.f.
d.o.f.
Lagrange multiplier d.o.f.

Intermediate vertical velocity

MODEL
VALIDATION

We can write it as an equation

for
only!

MODEL
VALIDATION

Hydrodynamic pressure computation

Hydrostatic correction

OUTCOME

OUTCOME

Convergence tests
* Indicates the elementby-element vector
product
IMA 2011

IMA 2011

13

A differential interpretation of the hydrostatic step

Coupling the two dynamics

The force acting on the boat is computed as

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

where

Let

s. t.

is the drag force, estimated by empirical formulae

is the solution of
is computed from a parametric description of the boat
displaced according to the computed rigid motion.

MODEL
VALIDATION

MODEL
VALIDATION

We use a simple explicit (staggered) scheme for the

interaction problem:

We can formally write the following equation for only

OUTCOME

OUTCOME

Fluid solution

Fluid solution
Boat
dynamics

IMA 2011

IMA 2011

A Final Touch

Example of grid on the xy plane an of a boat geometry

Most recent simulations have been carried out rewriting the

flow equation in a reference frame with the origin fixed in
the xy plane: smaller computational domain. It suffices to
add the term
to the momentum equation

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

At the far field we have implemented a first order linearized

radiation condition to reduce unphysical reflections
MODEL
VALIDATION

MODEL
VALIDATION

Different numerical schemes of predictor/corrector type have

been implemented for the boat dynamics

OUTCOME

IMA 2011

OUTCOME

IMA 2011

14

First test case: wave produced by a wigley hull moving at

constant speed (hydrostatic approximation)

Sinking and pitching motion

Ellipsoid with the following

characteristics

Froude number 0.316

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Sinking motion: return to equilibrium after a vertical

displacement
MODEL
VALIDATION

MODEL
VALIDATION

Comparison with the

theoretical Mach line

OUTCOME

OUTCOME

3940 N

IMA 2011

IMA 2011

Pitching motion

Wave hitting a boat

The boat returns to equilibrium after an angular displacement

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

OUTCOME

IMA 2011

IMA 2011

15

Coupling with the full dynamics (hydrostatic)

Hydrostatic vs hydrodynamic results

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

MODEL
VALIDATION

MODEL
VALIDATION

OUTCOME

OUTCOME

IMA 2011

IMA 2011

Conclusions

Acknowledgements

We have presented two reduced models for the dynamic of a

rowing boat which are able to provide reasonable fast answer
to designers and trainers

GEOMETRIC
PREPROCSSING

GEOMETRIC
PREPROCSSING

Ongoing and future work

for both models

Validation with experiments in collaboration with Filippi

Lido and the University of Ferrara

MODEL
VALIDATION

for the inequality constraint model

Integration with real boat geometries from CAD data

Coupling with a model of boundary layer to account for
friction

OUTCOME

IMA 2011

Filippi Lido s.r.l. for financial support and in particular Ing. Alessandro
Placido who introduced us to the wonderful world of rowing
Andrea Paradiso and Michele Altieri for the availability of some results
from their master thesis
Fausto Saleri for his important contribution to the original idea

MODEL
VALIDATION

OUTCOME

The work has been partially supported by a PRIN07 project of the Italian
MIUR
IMA 2011

16

References
E. Miglio, A. Quarteroni and F. Saleri. Finite element approximation of quasi-3D
shallow water equations. Com. Meth. Appl. Mech. Engng. 174(3-4):335-369, 1999
P. Causin, E. Miglio and F. Saleri. Algebraic factorizations for 3D non hydrostatic free
surface flows, Comp. Vis. Sci. 5(2):85-92, 2002

GEOMETRIC
PREPROCSSING

A. Mola. Models for olympic rowing boats. PhD Thesis, Politecnico di Milano, 2009
A. Mola. Multiphysics and multilevel fidelity modelling and analysis of olympic rowing
boat dynamics. PhD Thesis, Virginia Tech, 2010
L. Formaggia, E. Miglio, A. Mola and A. Montano, A model for the dynamics of rowing
MODEL
boats. Int. J. Numer. Meth. Fluids, 61(2):119-143, 2009
VALIDATION
L. Formaggia, E. Miglio, A. Mola and A. Scotti. Numerical simulation of the dynamics
of boats by a variational inequality approach. In Variational analysis and aerospace
engineering. 213-227, Springer, 2009
L. Formaggia, A. Mola. N. Parolini and M. Pischiutta. A three-dimensional model for
OUTCOME
the dynamics and hydrodynamics of rowing boats. Journal of Sport Engineering
and Technology, 224(1):51-61, 2010.
L. Tamellini, L. Formaggia, E.Miglio and A. Scotti. An Uzawa iterative scheme for the
simulation of floating boats. Submitted 2010.
IMA 2011

17