You are on page 1of 6

25th Numerical Towing Tank Symposium (NuTTS)

15-17 October 2023 - Ericeira, Portugal

FSI Analyses of a High-Performance Solar Boat Composite T-hydrofoil


Tomas L. T. M. Carreira∗ , Leigh Stuart Sutherland† , and Jose Manuel C. Pereira∗
∗ IDMEC-IST, Lisbon/Portugal, † CENTEC-IST, Lisbon/Portugal, ∗ LASEF-IST, Lisbon/Portugal,
tomasltmcarreira@tecnico.ulisboa.pt

1 Introduction
Hydrofoils were inadvertently discovered in 1861 leading to early experimentation around the turn of
that century and later widespread use in both military and passenger vessels. In recent years the use of
very low specific stiffness and strength CFRP materials has enabled an explosion in hydro-foiling sail
and power leisure craft, as well as a renewed interest in transport vessels. However, the complex nature of
these materials (e.g. anisotropy, multiple and complex failure modes, dependence on production process
and quality, variable mechanical properties, susceptibility to flaws etc.) together with a lack of in-service
design data, mean that tools and expertise to predict the behaviour of these foils are still under developed
and/or proprietary. Here, the aim is to develop a fluid-structure interaction (FSI) analyses of the composite
T-hydrofoil of the (Instituto Superior) Tecnico Solar Boat (TSB) S R03 prototype (Figure 1) designed for
the annual Monaco Energy Boat Challenge (MEBC).

(a) Madeira Solar Race championship (b) Hydrofoil wing and strut

Fig. 1: TSB SR03 6m solar-powered hydrofoil dinghy

The SR03 has a 1.5kWh battery, 6m2 of solar panels and a 10kW electric motor. Take off and cruise
speeds are 10 and 15 knots, respectively, the later for up to 3 hours. The angles of attack of the three very
stiff monolithic carbon-fibre-epoxy pre-preg T hydrofoils are independently adjusted by an integrated
electronic controller via data from an ultrasound bow sensor bow.
2 Background
Fluid-structure interaction is the reciprocal influence of a solid structure on a flowing fluid (Borst et al.,
2013). For hydrofoils, traditional approaches only modelling the wing as a rigid body may not be suffi-
cient since they fail to capture the effects of the structure deformation on the flow field, as verified for
windsurfing fins (vertical hydrofoils) by the work of Sutherland and co-workers, (Sutherland, 1993) up
to (Brito et al., 2022), who also investigated the bend-twist coupling behaviour of the fin under sailing
loads. (Zarruk et al., 2014), (Pernod et al., 2017), (Young et al., 2018), (Liao et al., 2019), (Giovannetti
et al., 2022), (Ducoin et al., 2009) and (Delafin et al., 2014) also studied various aspects of FSI and
bend-twist coupling of hydrofoils. This work aims to extend the work of Sutherland and co-workers to
evaluate if FSI and/or bend-twist coupling may be used to assist the design and optimisation of the TSB
SR03 hydrofoils and provides a current and constantly evolving snapshot of the ongoing M.Sc. thesis
work of the first author.
3 Structural model
The structural part of this FSI model is developed and calibrated using Siemens NX. The wing (Figure
1b) is of wingspan 800mm and root-chord 100mm, it is without sweep and the leading edge is elliptically
curved with minor axis at 80% span. The trailing edge is faired into the root. using symmetry, only half
the geometry is modelled in the computational environment to help keep computational costs reasonable.

1
The hydrofoil section is the NACA 631 412 (12% thickness-chord ratio ) laminar, cambered airfoil, often
used for low-to-moderate Reynolds number applications (Redesign ≈ 3000000), with a high Li f t/Drag
ratio when in operating points within the drag-bucket (Cl ∈ [0.3, 0.5]). The highly curved lower pressure
surface is a significant manufacturing challenge.
The wing layup has 52 layers of carbon fibre pre-impregnated in a low temperature cure epoxy resin.
The fibres are woven twill fabric, and plies become progressively shorter in order to fit in the mould space.
Ply mechanical properties were estimated using the Siemens NX improved woven ply material model
(Table 1) and were then validated against equivalent material properties from classical micromechanical
models proposed in (Vignoli et al., 2019).

Symbol Description Value Units


ρc Density 1.384 kg/m3
t Ply finished thickness 0.23 mm
E xx , Eyy 58.26 GPa
Young’s Modulus
Ezz 3.90 GPa
ν xy 0.03248 -
Poisson’s ratio
ν xz , νyz 0.2955 -
G xy 3.245 GPa
G xz Shear Modulus 1.478 GPa
Gyz 1.794 GPa

Table 1: Ply equivalent mechanical properties computed using Siemens NX improved woven model

Discretisation was achieved using the native mesher from NX. Given the quasi-single curvature of
the leading and trailing edge regions, 2DCQUAD8 shell elements were used (Hexagon, 2022). Face
partitions were made to accommodate the varying ply-schedule over the wing surface according to the
available manufacturing templates. In the coupled analysis the strut geometry is fixed, thus all degrees
of freedom of the grid-nodes on the strut interface were fixed for all studies. Spanwise and vertical rota-
tions, as well as spanwise translation were fixed at the root airfoil nodes to enforce symmetry boundary
conditions. Since the upper and lower surface meshes are connected only at the leading edge nodes, per-
fect bonding glue conditions were applied at the trailing edge, at a fixed distance of 0.1mm and at the
negative-normal element faces in the wing surfaces.
3.1 Solution Verification and Validation
A verification study was performed to guarantee a mesh independent structural solution. Successively
refined unstructured meshes were generated for the wing geometry and discretization errors were esti-
mated, in a least squares sense, using the procedure proposed by (Eca and Hoekstra, 2014).

(a) Wing-tip vertical displacement (b) Elastic energy

Fig. 2: Discretization uncertainties considering grid refinement ratio based on RMS volume

Two solution quantities were considered, elastic strain energy (U), a global quantity, and maximum
vertical tip displacement, δytip , a local output of interest. In 2, the solution values lie within a narrow
range for all meshes and uncertainty values suffer little change among the three most refined meshes,

2
the third mesh was chosen as the best compromise between low estimated uncertainty and reasonable
computational resources, with a target cell size of h = 1mm and Uu = 17% and Uδy = 8.4% uncertainties,
respectively.
Validation of the structural model with servo-hydraulic press tests was attempted through the course
of this work. However, the lack of trustworthy information about the exact layup schedule was an insur-
mountable problem. The layup used for the development of the FEM model was considered as a baseline
for all the studies performed.
4 Hydrodynamic Model

The hydrofoil is designed for July Mediterranean conditions, with a density of ρ = 1025kg/m3 and
dynamic viscosity of µ = 0.00109Pa · s.(SeaTemperatures.net, 2023) The fluid is assumed to be incom-
pressible and inflow velocity constant.
The same idealized geometry described in section 3 was used for the fluid analysis and the mean
aerodynamic chord (MAC) of the wing, approximately 68mm, was considered for the reference geomet-
ric length of the problem. A prismatic domain was used to model the fluid continua with the inlet and
top and bottom walls at 10 × MAC of the wing leading edge, the outlet at 15 × MAC and the side wall
at 12 × MAC (or 3 × S emispan). The boundary conditions used were: Velocity Inlet for inlet, top and
bottom surfaces, Symmetry Plane for side and symmetry plane surfaces, Pressure Outlet for outlet and
No-Slip Wall for wing and strut geometries. Unstructured meshes were generated with STAR-CCM+ na-
tive algorithms: Automated Mesh, Surface Remesher, Polyhedral Mesher and Advancing Layer Mesher.
The Navier-Stokes equations in their Reynolds Averaged Navier-Stokes (RANS) simplified formu-
lation were discretized using the finite volume method. K − Ω S S T eddy viscosity model was used
to solve mean flow quantities transport equations, providing closure to the RANS equations. Given the
low Reynolds Number for the reference length (MAC Re MAC = 508484), a laminar boundary layer is
expected over much of the chordwise surface. Separation-induced transition caused by the adverse pres-
sure gradient aft to the maximum thickness of the wing is expected as is reattachment of the laminar
boundary layer due to increased mixing ability of the turbulent flow. The phenomena described above
is called a laminar separation bubble and has been observed in similar Reynolds Number applications
(Delafin et al., 2014)(Ducoin et al., 2009)(Young et al., 2018). STAR-CCM+’s native correlation-based
two transport equation γ − Reθ model was used to predict the onset of transition on the wing surfaces,
thus capturing laminar separation bubbles and their effect on the pressure and shear hydrodynamic loads.
The Implicit Coupled-Flow is used with a second-order upwind discretization scheme for convective and
diffusive fluxes on all transport equations. Given the inherent transient nature of the flow, an implicit
second-order backward scheme was used for time integration. However, a pseudo-transient approach
was used to decrease computational cost and obtain relevant steady state hydrodynamic loads on the
composite structure.
5 Fluid-Structure Interaction

Coupled analyses were performed using the native features of the Simcenter STAR-CCM+ and Siemens
NX/Simcenter NASTRAN softwares.
5.1 One-way FSI analyses
Initially, data from CFD simulations was imported to NX using STAR-CCM+’s Surface Data Mapper
feature. Loads were computed assuming that the wing and strut geometries behave as rigid bodies. This
is a very low computational cost route, but the usefulness of the results will depend on finding low wing
tip bending and twist displacements.
5.2 Two-way FSI analyses
The far more computationally expensive approach of a coupled analysis between fluid and structural
solvers was then undertaken. A pseudo-transient explicitly or weakly coupled approach was considered
since the coupling algorithm used for the exchange of information between the STAR-CCM+ fluid solver
and the Simcenter NASTRAN structural solver only allows data exchange once per time-step (STAR-

3
CCM+, 2022) meaning that convergence of the fluid solver is not verified within each time-step (Benra
et al., 2011). This approach could be less computationally expensive, but to reach a stable solution a
smaller time-step is often needed (Giovannetti, 2017).
6 Parametric Studies
6.1 Take-off and Cruise structural behaviour
The two most critical operating conditions were considered: Cruise configuration where drag is at its
minimum to allow for minimum required power and maximum flight duration, and Take-off configuration
where wing lift must be maximum. Actual cruise operating conditions with free-stream velocity, U∞ =
8m/s and wing angle of attack AoA = 0◦ were used. Figure 3 and Table 2 show the deformed wing
shape and the load and deflection results, respectively, under cruise conditions computed using a one-
way analysis.

Fig. 3: Deformed shape of the wing in cruise configuration- one-way FS I

Variable Description Value Units


CL Lift coefficient 0.0878 -
CD Drag coefficient 0.0129 -
δytip Tip vertical displacement 4.196 mm
θztip Tip twist -0.9203 ◦

θ xmax Maximum bending slope 1.082 ◦

Table 2: One-way FS I results for cruise configuration: U∞ = 8m/s and AoA = 0◦

Tip vertical displacement is only around 1% of the wing semi-span, indicating that the wing is be-
having very rigidly, and hence suggesting that performing a two-way coupled analysis may well be un-
justified. However, integrated coupled analysis is currently being completed in order to formally verify
this.
Negative or wash-out tip twist (following the positive as nose-up convention) was observed at cruis-
ing speed. At this point in the analyses it is only possible to hypothesize how this behaviour can affect
the wing performance, but it is possible that this behaviour helps reduce local lift at the wing-tips, thus
reducing span-wise flow and the resulting induced drag. Analyses of the take-off structural behaviour is
currently still under way.
6.2 Interface separation impact on lift
Whilst performing an initial 2D verification study, lift force coefficient convergence issues were encoun-
tered, associated with high turbulent kinetic energy (k) fluctuations and high residuals in the neighbouring
region to the T-junction wing-strut interface. A small separation region was expected in this region, but
the streamlines of 4a show that this is not confined to the near wake of the interface, but also extends
further upstream to the lower strut and onto the upper surface. The chord/wise upper surface wall shear
stress contour also shown in 4b (only negative values shown for clarity), indicates that the expected
laminar separation bubble (visible further along the span) was disrupted by the presence of a separation
region.
In Figure 4b, the separation (red) and reattachment (blue) lines show an upstream deflection of the
laminar separation bubble region near the wing-strut interface, and at around mid-chord next to the in-
terface a channel between reattachment lines is evident. Together with the negative span-wise wall shear
stress (moving from root to tip) this channel suggests a deflection of chord-wise flow in the spanwise
direction, towards the tip. This is expected to cause a significant reduction in lift coefficient. Hence, a
comparative study featuring several simpler estimates for the lift coefficient of the wing at cruise con-
ditions, U∞ = 8m/s and AoA = 0◦ was performed. Estimates were made through two approaches: The

4
(b) Separation(red) + Reattachment(blue)
(a) Streamlines + Wall shear stress [x] lines + Wall shear stress [z]

Fig. 4: Evidence of flow ’channelling’ and separation expansion in wing’s extrados

simulation of the wing and strut geometry using the low-fidelity code XFLR5, and an estimate of the 3D
solution using available 2D solutions (from the verification study and XFOIL) and the simple Prandtl
equation (Cantwell, 2014).
!
AR
CL 3D
= Cl 2D
(1)
AR + 2
Where ARis the aspect ratio of the wing.

Solution Cl2D CL3D CL3D f ull/i [%]


2D veri f ication 0.2830 0.2421 36.26%
XFOIL 0.3402 0.2910 30.16%
XFLR5 - 0.2893 30.34%
3D f ull - 0.0878 -

Table 3: Comparative study of lift coefficients at cruise conditions (U∞ = 8m/s and AoA = 0◦ )

The results shown in table 3 indicate that the FSI lift coefficient was only around 30% of these
simpler estimates that do not consider any effects of separation bubbles nor the wing-strut interface,
underscoring the substantial impact on lift of these two separation regions. Practically, This results in
a substantial under-performance of the hydrofoil wing compared with the design case (which used the
simpler methods) at cruise operating conditions, which would force the controller to increase the angle
of attack resulting in drag increase and hence a substantial efficiency penalty.
6.3 Composite layup and bend-twist coupling
The layup used for the actual wing was considered as the baseline for a bending-twist coupling parametric
analysis (currently under way) for possible incorporation of passive control mechanisms into the design.
Various plies are rotated in-plane by various amounts and the effect on tip twist via changes in bend-twist
coupling measured. Any associated changes in bending, tip displacement are also recorded. Finally, work
is also underway to explore the effects of using a more refined lay-up schedule with more 0◦ , +45◦ and
−45◦ and less 90◦ fibres.
7 Conclusions
This work considers a racing solar electric boat monolithic carbon-fibre epoxy T −hydrofoil. Structural
finite element analysis and hydrodynamic models have been developed in Siemens NX and Simcenter
STAR-CCM+. FEA and CFD models were coupled in a one-way model and development of an fully
coupled hydro-elastic model is under way. Laminar separation bubbles and a separation region at the
wing-strut interface significantly increased both the model complexity and computational resources, and
importantly suggested very simply implemented (via a simple cowling) design changes to significantly
improve the efficiency of the craft. The initial one-way analyses indicated a very stiff structural behaviour
of the wing with minor tip deflection and a reasonable twist behaviour. Hence, simpler and much cheaper

5
one-way FSI analyses appear to be sufficient for this case, but nonetheless a fully coupled model is
currently under development to verify this.
The findings made throughout the course of this study, up to the present date, support the relevance
of high-fidelity computational analysis on the process of designing a hydrofoil system, albeit not being a
substitute for experimental testing and prototyping, but rather a useful complement.
8 Acknowledgements
We thank the Tecnico Solar Boat (T S B) team for providing the hydrofoil wing and strut prototypes that
were featured as a case study in this work. We thank the Department of Mechanical Engineering (DEM)
of Instituto Superior Tecnico for providing the necessary computational resources for the development
of this work. We also thank LAS EF for providing Simcenter STAR-CCM+ licenses.
This work is financed by national funds through the FCT −Foundation for Science and Technology,
I.P., within the scope of the project UIDB/50022/2020.
References
Benra, F. K., Dohmen, H. J., Pei, J., Schuster, S., and Wan, B. (2011). A comparison of one-way and
two-way coupling methods for numerical analysis of fluid-structure interactions. Journal of Applied
Mathematics, 2011.
Borst, R. D., Nithiarasu, P., Tezduyar, T. E., Yagawa, G., and Zohdi, T. (2013). Computational Fluid-
Structure Interaction: Methods and Applications. John Wiley and Sons Ltd.
Brito, M. C., Sutherland, L. S., Pereira, J. M. C., and Arruda, M. R. (2022). Fluid-structure interaction
analyses for hydro-elastic tailoring of a windsurfer fin. Journal of Marine Science and Engineering,
10.
Cantwell, B. J. (2014). Chapter 12 wings of finite span.
Delafin, P. L., Deniset, F., and Astolfi, J. A. (2014). Effect of the laminar separation bubble induced
transition on the hydrodynamic performance of a hydrofoil. European Journal of Mechanics, B/Fluids,
46:190–200.
Ducoin, A., Deniset, F., and Astolfi, J.-A. (2009). Computational and experimental investigation of flow
over a transient pitching hydrofoil. European Journal of Mechanics, 28:728–743.
Eca, L. and Hoekstra, M. (2014). A procedure for the estimation of the numerical uncertainty of cfd
calculations based on grid refinement studies. Journal of Computational Physics, 262:104–130.
Giovannetti, L. M. (2017). Fluid structure interaction testing, modelling and development of passive
adaptive composite foils.
Giovannetti, L. M., Farousi, A., Ebbesson, F., Thollot, A., Shiri, A., and Eslamdoost, A. (2022). Fluid-
structure interaction of a foiling craft. Journal of Marine Science and Engineering, 10.
Hexagon (2022). MSC Nastran 2022.1 Linear Static Analysis User Guide. Hexagon AB, Munich,
Germany, 2022.1 edition.
Liao, Y., Garg, N., Martins, J. R., and Young, Y. L. (2019). Viscous fluid structure interaction response
of composite hydrofoils. Composite Structures, 212:571–585.
Pernod, L., Ducoin, A., Sourne, H. L., and Sigrist, J.-F. (2017). Coupled numerical simulation of an
aluminum and a composite hydrofoil in steady flows.
SeaTemperatures.net (2023). Monte carlo sea temperature.
STAR-CCM+, S. (2022). Simcenter STAR-CCM+ 2302 User Guide. Siemens, 2302 edition.
Sutherland, L. S. (1993). University of southampton windsurfer fin hydrodynamics.
Vignoli, L. L., Savi, M. A., Pacheco, P. M., and Kalamkarov, A. L. (2019). Comparative analysis of
micromechanical models for the elastic composite laminae. Composites Part B: Engineering, 174.
Young, Y. L., Garg, N., Brandner, P. A., Pearce, B. W., Butler, D., Clarke, D., and Phillips, A. W. (2018).
Load-dependent bend-twist coupling effects on the steady-state hydroelastic response of composite
hydrofoils. Composite Structures, 189:398–418.
Zarruk, G. A., Brandner, P. A., Pearce, B. W., and Phillips, A. W. (2014). Experimental study of the
steady fluid-structure interaction of flexible hydrofoils. Journal of Fluids and Structures, 51:326–343.

You might also like