You are on page 1of 16

ARTICLE IN PRESS

International Journal of Impact Engineering 31 (2005) 929–944


www.elsevier.com/locate/ijimpeng

A survey of numerical models for hail impact analysis using


explicit finite element codes
Marco Anghileri, Luigi-M. L. Castelletti, Fabio Invernizzi, Marco Mascheroni
Dipartimento di Ingegneria Aerospaziale, Politecnico di Milano, Via La Masa, 34–20156, Milano, Italy
Received 5 April 2004; received in revised form 10 June 2004; accepted 12 June 2004
Available online 8 October 2004

Abstract

Hailstone impact is an actual threat for the integrity of aircraft structures such as leading edges, and
forward sections. Though the analysis of weather conditions reduces the occurrence of intersections
between flight routes and hailstorm regions, sometimes the passage through a hailstorm becomes inevitable
and, in such a case, it is mandatory that the aircraft structures show an appropriate level of tolerance to
damages caused by a hail impact. Therefore, as experimental tests are both expensive and troublesome, it is
important to develop numerical models, which eventually support the design of high-efficient and hail-
proof structures. Accordingly, in the present research, using LSTC LS-Dyna, three numerical models of
hailstone have been developed: finite element, arbitrary Lagrangian Eulerian, and smoothed particle
hydrodynamics model. Initially, these three models had been validated referring to a documented
experimental test and, subsequently, used to reproduce the impact of a hailstone with the nose-lip of a
nacelle intake. Advantages and disadvantages of the three hail models have been evaluated and it has been
concluded that the smoothed particle hydrodynamics model is the most efficient and effective for the
analysis of the event.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Airworthiness; Hail impact; Explicit finite element codes

Corresponding author. LAST Labs, Politecnico di Milano, Edificio G, Via Durando, 10–20158, Milano, Italy. Tel.:
+39-02-2399-7155; fax: +39-02-2399-7153.
E-mail address: luigi.castelletti@polimi.it (L.-M.L. Castelletti).

0734-743X/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijimpeng.2004.06.009
ARTICLE IN PRESS

930 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

1. Introduction

Hailstone impacts represent an actual threat for aircraft structures. Indeed, it is not unlikely
that aircraft, flying in adverse weather conditions, pass through hailstorm regions. Even a short
while in these regions, is sufficient to cause serious damage to aircraft structures such as abrasions,
dents and, in some cases, perforations.
The most exposed parts in this event are the leading edges of wings and tail control surfaces, the
fuselage forward sections, the engine nacelles, and other accessories such as the radar antenna and
the landing lights [1]. By the observation of damaged aircraft and by specific experimental tests, it
has been recognised that the extent of the mentioned damage depends on the features of both the
hailstone (mass, impact angle and velocity), and the impacted structures (geometry and material).
The analysis of weather conditions has so far made it possible to avoid intersections between
flight routes and hailstorms regions. Nevertheless, sometimes the passage through a hailstorm
becomes inevitable. Therefore, the structures of modern aircraft are called by specific requirements
(such as those in the European JAR) to guarantee a certain level of functionality after having been
impacted by a number of hailstones. Indeed, though hail impact has been recognised as a serious
problem since the early 1950s [1], recent studies concerning the consequences of the hail impact with
aircraft structures are somewhat rare. Since experimental tests are expensive and difficult to
perform, it is straightforward to understand the importance of developing numerical models to
reproduce and analyse the consequences of hailstone impacts. Indeed, after being properly validated
also referring to experimental evidences, these models might represent a powerful (efficient and
effective) tool to design high-performances and hail-proof structures.
In particular, the aim of this work has been to develop a numerical model of hailstone to
simulate the hail impact against aircraft structures by means of a general-purpose (commercially
available) explicit Finite Element (FE) code: LSTC LS-Dyna [2,3]. In particular, three different
models have been investigated: the customary Lagrangian FE model, the Arbitrary Lagrangian
Eulerian (ALE) model, and a model based on Smoothed Particle Hydrodynamics (SPH) method.
Initially, these models have been validated referring to the results of experimental tests carried out
by the British Royal Aircraft Establishment (RAE), and lately also cited in NASA technical notes
[4,5]. Then, the mentioned models have been used to reproduce the damages caused by a hailstone
onto the nose-lip of a turbofan engine intake. In particular, to verify the reasonability of the
obtained numerical results, these were (qualitatively) compared with the photographic
documentation of actual hailstone impacts occurred to an airlines aircraft. As a result, it has
been possible to highlight the advantages and disadvantages of the three different hail models.
Finally, the same numerical models have been also used to verify that the structure of the intake
was such to avoid hailstone penetration inside the nacelle airframe when considering the impact
velocity required for certification (such as JAR-E 970 and ACJ E 970).
Concluding, the feasibility of the different models for simultaneous impact of a number of
hailstones has been considered—and this is likely to be a further development of the present work.

2. Numerical model of the hail

As already mentioned, in this research, to reproduce a hail impact, three different numerical
hailstone models (Lagrangian FE, ALE and SPH) have been validated referring to an actual
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 931

experimental test carried out by the British RAE (and described in the NASA technical note D-
6102).

2.1. Experimental test carried out by British RAE

The test considered consisted of the impact of a 25.40 mm diameter hailstone against an
aluminium alloy 2014-T4 plate (Fig. 1) with an initial impact velocity of 192 m/s. The plate was a
square 0.91 mm thick panel of 305 mm side-edge, which was fixed at the boundaries with blind
rivets so that eventually the free target surface consisted of a square region of 200 mm side-edge.
The plastic strain observed after the impact (measured with regard to Section A–A in Fig. 1)
was used in the following as reference (the continue line in Fig. 2). In particular, the maximum
displacement (measured in the centre of the panel) was 11.20 mm.

2.2. Lagrangian finite element model of the hailstone

The Lagrangian FE approach is typical for continuum mechanics [6] and strongly
recommendable for the analysis of impact events. This approach is extremely efficient when
considering nonlinear problems, though it has its weak point in the excessive mesh distortions—
which are rather typical in events featuring soft-bodies or fluid-like materials such as in the case
considered.
Following a Lagrangian approach, the FE mesh is constructed on the material and therefore,
when the material undergoes large distortions, also the mesh undergoes the same large distortions,
which cause an unacceptable loss in accuracy, a considerable increase in required CPU time (due
to the fall of time-step value), and sometimes, a premature analysis termination.

Fig. 1. Experimental test panel.


ARTICLE IN PRESS

932 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

The FE model of the hailstone reproduced its actual shape, by approximating a 12.7 mm radius
sphere with hexahedral solid elements. The common eight-node three-linear isoparametric solid
elements were used and, as a result of a specific sensitivity analysis, the mesh consisted of 3045
elements, with a reference characteristic length of 1.2 mm.
Moving from an established Constitutive Law [7,8], the ice was modelled as an elastic plastic
material with failure (*MAT 13 of LS-Dyna [2,3]). Indeed, this material model allows a plastic
hardening behaviour that adequately reproduces the effect of the propagation of the micro cracks
inside the ice before it crushes, reaching a fluid-like state. When the plastic failure strain is
reached, all shear stress components are relaxed to zero. Furthermore, if the tensile failure
pressure is reached, the material carries only hydrostatic compressive stresses as a fluid. The
mechanical properties of the ice are summarised in Table 1.
Obviously, as already noticed in [7,8], this is only a simplified representation of a complex
material such as the ice. In fact, this model does not consider the complicated nonlinear mutual
dependency between pressure and volume and the influence of the strain rate on the failure
strength. However, it was showed [7,8] that this material model makes it possible to reproduce
correctly the ice behaviour during a high-velocity impact and therefore it has been adopted.

2.3. Arbitrary Lagrangian Eulerian model of the hailstone

The ALE approach is meant to combine the advantages of both Lagrangian and Eulerian
approaches. Following a Eulerian approach, the material flows through a mesh fixed in the space.
The ALE approach differs in that the Eulerian mesh can move arbitrarily and, in case,
independently on the motion of the material. Consequently, it has an advantage over the pure
Eulerian approach when the motion of the material covers a wide region of the space. In this case,
the number of elements using the Eulerian approach should be so large to maintain a reasonable
accuracy in the calculation, that the CPU time required for the analysis would be unacceptable.
With regard to the ALE model of the hailstone, it is worth noticing that it was necessary to
discretise not only the hailstone, but also a small surrounding region. In this way, it was possible
to avoid the outflow of the hailstone from the Eulerian mesh due to the velocity of the material,
which was higher than the expansion-translation velocity of the Eulerian mesh.
The ALE mesh of the hailstone and the surrounding void region consisted of 5888 hexahedral
elements. For these elements, it was adopted the standard ALE formulation with one integration

Table 1
Mechanical properties of the hailstone used for *MAT_13 of LS-Dyna [2,3]

Properties Values

Density (kg/m3) 846


Elastic shear modulus (GPa) 3.46
Yield strength (MPa) 10.30
Hardening modulus (GPa) 6.89
Bulk modulus (GPa) 8.99
Plastic failure strain 0.35
Tensile failure pressure (MPa) 4.00
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 933

point [3]. The Eulerian mesh was prescribed to move following the hailstone mass weighted
average velocity [3].
In order to characterise the ice behaviour, the same material model used for the Lagrangian FE
model (Table 1) was adopted.

2.4. Smoothed particle hydrodynamics model of the hailstone

The SPH method was initially developed (1997) to study astrophysical and cosmological
phenomena, subsequently, it was extended to typical problems of computational fluid dynamic
(CFD) and finally to continuum mechanic.
The main difference between SPH and FE method regards the discretisation of the continuum.
The SPH is a meshless method: the mesh is replaced with a set of particles endowed with a mass.
No direct connectivity exists among the particles: the particles are the basis of an interpolatory
scheme based on the kernel function that is the core of the method.
The SPH model of the hailstone was developed starting from both the FE model previously
described, and the real physical and mechanical hail properties [9]. An extremely regular
distribution of the particles was provided to improve the accuracy in the solution. In particular,
the distance between the particles was chosen equal to the value of the characteristic length of the
previously described FE model. Accordingly, the model consisted of 4169 particles.
Since the material model used for the FE model is not implemented for the SPH solver, it was
necessary to use a different Constitutive Law [10]. Several simulations were performed: a number
of material models were investigated. For each one of these material models, a number of values
of the characteristic parameters were considered. The obtained results were compared with
experimental data [4,5]. As a result, a SPH model able to provide a good numerical-experimental
correlation was obtained [10].
The elastic plastic hydrodynamic material model (*MAT 10 of LS-Dyna [2,3]) was used for ice,
and the adopted mechanical properties are provided in Table 2. This model has made it possible
to represent the hailstone behaviour correctly both in the early stages of the impact, when it is
characterised by a high stiffness, and in the subsequent stages, when, cracked after impact, it
behaves like a fluid.
The material model is characterised by a failure criterion relative to the tensile stress which
works in such a way that when the tensile failure stress is reached, the deviatoric stresses
component is set to zero and the material can sustain only compressive stresses. The plastic failure

Table 2
Mechanical properties of the hailstone used for *MAT_10 of LS-Dyna [2,3]

Properties Values

Density (kg/m3) 846


Elastic shear modulus (GPa) 3.46
Yield strength (MPa) 10.30
Plastic hardening modulus (GPa) 6.89
Pressure cut-off (MPa) 4.00
ARTICLE IN PRESS

934 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

strain criterion was not defined because it was considered unsuitable to reproduce the failure
mechanisms of ice correctly. However, the tension instability, typical of SPH method, partially
supplied this lack. As the adopted material model requires an Equation of State (EOS), the
polynomial EOS of the water was used [11].

2.5. Comparisons and remarks on the different hail models

The results obtained after the simulations carried out using the three different hail models were
compared qualitatively and quantitatively among themselves and with the experimental test data
[4,5].
Considering the graphical evidence of the impact (Fig. 2), it is evident that FE mesh (Fig. 2a)
undergoes large distortions and, therefore, it is straightforward to conclude that the use of the FE
model seems feasible only in the early stages of the impact. On the contrary, the ALE model (Fig.
2b) gives a somewhat accurate description of the cracked hailstone also when it is characterised by
a fluid-like state. Nevertheless, it is the SPH model (Fig. 2c) that reproduces the hailstone
behaviour, visually, in a way closer to common experience.

5
Deflection [mm]

0
-5
-10
-15
100 110 120 130 140 150 160 170 180 190 200
Position [mm]

(a) Experimental Data Numerical Results (FE model)

5
Deflection [mm]

0
-5
-10
-15
100 110 120 130 140 150 160 170 180 190 200
Position [mm]

(b) Experimental Data Numerical Results (ALE model)

5
Deflection [mm]

0
-5
-10
-15
100 110 120 130 140 150 160 170 180 190 200
Position [mm]
(c) Experimental Data Numerical Results (SPH model)

3
Fig. 2. Hail impact at t=0.150  10 s and numerical-experimental correlation. (a) FE; (b) ALE; and (c) SPH models.
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 935

The quantitative comparisons (Fig. 2) were made referring to the plastic strain of the plate
measured after the experimental tests [4,5]. In Table 3 the maximum plastic strain in the centre of
the panel is directly compared with the displacement measured in the test. In particular, it is worth
noticing that the FE and SPH models give errors within 1%, the ALE model within 4%.
The required CPU time was also monitored. Considering the FE model of the hailstone, a
drastic drop in the time-step was observed after the impact (due to the large mesh distortion).
Indeed, the time-step decreases also in the SPH and ALE model, but definitively less (Table 4).
Using a common PC (a Pentium 4–1700 MHz CPU, and 256 Mb RAM), the required time for
0.7 ms real-time simulation ranged from a minimum of about half an hour (SPH model) to a
maximum of about 18 h (FE model)—as reported in Table 4.

3. Impact of a hailstone against a turbofan engine intake

The intake of a turbofan engine nacelle is one of the aircraft parts more seriously damaged
when the aircraft cross a hailstorm region. Though this event is not likely to cause an air disaster,
the penetration of a hailstone (at high velocity) inside the intake structure is likely to cause the loss
of the engine. In fact, the electronic system that controls the engine is usually positioned inside the
intake airframe. Therefore, it makes sense to analyse the consequences of a hail impact against a
turbofan engine intake. In particular, in this research, considering the actual structure of a
turbofan engine intake, the dents caused by the developed hail models were initially compared
with the documented damages due to hail impacts against similar structures. Then, considering

Table 3
Numerical-experimental correlation

Maximum displacement (mm) Relative error (%)

Experimental test 11.20


Finite element model 11.25 0.4
Arbitrary Lagrangian Eulerian model 11.70 4.0
Smoothed particle hydrodynamics model 11.28 0.7

Table 4
Required CPU-time

Hail model Time-step CPU-timea

Initial (s) Final (s)


8 10
Finite element model 4.39  10 9.51  10 39 h, 9 min
8 8
Arbitrary Lagrangian Eulerian model 4.39  10 1.69  10 14 h, 30 min
7 7
Smoothed particle hydrodynamics model 2.07  10 1.85  10 1 h, 4 min
a
CPU-time required by a Pentium 4–1700 MHz CPU, and 256 MB RAM PC.
ARTICLE IN PRESS

936 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

the impact velocity prescribed for hail ingestion (JAR-E 970 and ACJ E 970), it was verified that
the structure was able to protect the mentioned system.

3.1. Numerical model of the nacelle intake

The FE model reproduced in detail the main features of an actual intake. In fact, it was built on
the geometry of the intake of a turbofan engine developed in a preliminary design phase. As a
reference on the intake dimensions, the inner diameter measured at the leading edge is 1.60 m.
The FE mesh (Fig. 3) consisted of 16 parts and 63013 four-nodes shell-elements. Five of these
parts are made of composite material and one, the inner barrel, adopting a sandwich technology
(composite materials and aluminium honeycomb)—typical in aircraft industry as it allows high
strength/weight ratios and a good noise control. Four parts are made of titanium alloy Ti-6Al-4V
and the remaining of two distinct aluminium alloys: Al 2219-T62 the nose-lip, and Al 2024-T6 the
other parts.
In order to obtain the necessary accuracy avoiding excessive computational efforts, the shell
elements had a relatively large reference length (10 mm) throughout the whole model but in the

Fig. 3. FE model of the intake: (a) full model and (b) a detail.
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 937

nose-lip region, where a locally finer mesh was built (2.5 mm). The Belytschko–Tsay formulation
(that is the default formulation for LSTC LS-Dyna) was adopted.
As a remark, it is worth noticing that the sandwich panels were not modelled in detail, but
equivalent shell elements were used. Such a model made it easier to change the thickness of the
honeycomb core, considering that, in a preliminary design phase, changes in thickness are
somewhat usual. Furthermore, it was observed that the sandwich panels were not directly
involved in the hail impact and, therefore, an accurate model of the part was not deemed
necessary.
The parts of the intake made up with metallic material were modelled using the elastic piecewise
(isotropic) linear plasticity material model (*MAT_24 of LS-Dyna [2,3]), using for the mechanical
properties of the materials customary values (also indicated in the MIL handbooks). Appropriate
Cowper–Symond coefficients were defined to model the effect of high strain rate deformation.
Also, for the nose-lip material, the residual stress was considered. The composite material used in
manufacturing the inner and outer barrel, a carbon fibre reinforced plastic (CFRP) woven with
resin volume fraction of 42%, was modelled with a specific material model (*MAT_54 of LS-
Dyna [2,3])—already validated [12].
Concluding, it is important to notice that the riveted junctions were not modelled also because,
since the FE model was developed on a preliminary design, the characteristics (number,
placement, and mechanical properties) of the riveted junctions were not available.
Contacts among the parts were carefully modelled. Furthermore, in order to reproduce the
actual constraint that the airframe of the engine imposes to the intake, the inner and outer flanges
of the aft part of the intake (in the FE model) were fixed.

3.2. Comparison with documented damages due to hail impacts

In order to verify the accuracy of the hail models developed in the first phase of the research,
initially, simulations reproducing the impact against a nacelle nose-lip were performed using
velocities close to those of documented accidents.
In particular, the accident occurred to the Douglas DC-9-81 (MD 81) of the Scandinavian
airlines system (SAS) was considered [13]. The aircraft, flying at a true-air speed (TAS) of about
140 m/s in the middle of a heavy hailstorm characterised by hailstones with diameters up to 5 mm,
reported serious dents in the nacelle leading edge. The largest dents were within 50 and 70 mm in
diameter and 5 mm deep (Fig. 5). In analogy with this accident, the impact of a hailstone at 140 m/
s was reproduced using the described FE model of a turbofan engine intake. This model does not
represent specifically the engine intake of a MD 81, but it reproduces the intake of an aircraft of
similar dimensions.
The three hail models previously investigated were used. Also, since the hailstones are characterised
by extremely various dimensions, to reproduce the most common conditions in which an aircraft can
incur, it was decided to perform simulations using two different dimensions for the radius of hailstone:
12.70 mm (the same previously used) and 21.35 mm (the dimension of a golf ball)—the model
of which was obtained from the former by scaling the dimensions. The most critical conditions were
considered: the hailstone impacted the most prominent point of the intake lower section, in a direction
parallel to the nacelle axis. As a result of the analyses, it was observed that in none of the considered
cases the hailstones caused the failure of the nose-lip and penetrated inside the intake structure
ARTICLE IN PRESS

938 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

(Fig. 4)—like in the documented case. Furthermore, it was noticed that the three hail models
produced similar dents quantitatively (Tables 5 and 6) and qualitatively (Fig. 4).
Despite the constructive differences and the lack of relevant data on the actual hail impacts, the
damages numerically obtained are definitively similar to the actual dents in dimension and in
appearance (Fig. 5).

3.3. Simulations considering the required actual impact velocity

After verifying the numerical model of the hail with regard to actual hail impacts, simulations
were performed with an impact velocity of 185 m/s—which, for the considered aircraft, is the hail
ingestion velocity indicated by the JAR/ACJ-E 970 requirements for certification of the intake.
When considering the impact of the 12.70 mm radius hailstone (Fig. 6), it is straightforward
to observe that the FE and SPH hailstone models produce similar results with respect to the
residual plastic strain. On the contrary, the impact of the ALE model of the hailstone is less

Fig. 4. Dents due to hailstones of 12.70 mm (left) and 21.35 mm (right) radius. (a) FE; (b) ALE; and (c) SPH models.
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 939

Table 5
Dents due to a 12.70 mm radius hailstone

Hail model Diameter (mm) Depth (mm)

Finite element model 38 3


Arbitrary Lagrangian Eulerian model 39 4
Smoothed particle hydrodynamics model 40 4

Table 6
Dents due to a 21.35 mm radius hailstone

Hail model Diameter (mm) Depth (mm)

Finite element model 118 10


Arbitrary Lagrangian Eulerian model 120 9
Smoothed particle hydrodynamics model 121 10

severe—so that, eventually, in contrast with the two other models, no collapses in the nacelle were
observed. This occurrence is somewhat typical in Lagrangian/Eulerian coupled analyses and,
basically, depends on the definition of the coupling between Lagrangian and Eulerian meshes.
The CPU-time required to run the simulations using the SPH model of hailstone was
considerably smaller than the runtime of the other two models—as shown in Table 7.
When considering the impact of the 21.35 mm radius hailstone (Fig. 7), the results showed that
the FE model of the hail is not suitable for the analysis of the event. In fact, after having impacted
the nose-lip and penetrated inside the airframe, the distortion of the hailstone mesh was such to
cause a premature termination of the simulation. As a remark, it is worth noticing that, in some
cases, an erosion criterion could be defined to avoid excessive mesh distortion. Nevertheless, it
was not used here because, to be effective, it produced damages, which are far from the evidences
collected in real hailstone impacts. Instead, using the ALE and SPH models of the hailstone, the
simulation reached a normal termination. In both cases, the hailstone, after breaking the nose-lip,
penetrated inside the airframe impacting the tube and the bulkhead.

3.4. Final remarks

Although all the developed hail models are appropriate for the analysis of the impact of
hailstone with compliant structures, the SPH model results in being the most suitable for the
impact with the intake of a turbofan engine. In fact, this model provides accurate solutions and
requires small CPU-time—these features make the SPH model a reliable and effective design tool.
Furthermore, this model has made it possible to analyse the penetration of the hailstone inside the
airframe and eventually it seems the only hail model suitable to analyse also the simultaneous
ARTICLE IN PRESS

940 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

Fig. 5. Dents on the nose-lip of turbofan engine intakes.


ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 941

Fig. 6. Impact of the 12.70 mm radius hailstone. (a) FE; (b) ALE; and (c) SPH models.

Table 7
Required CPU-time

Hail model Time-step CPU-timea

Initial (s) Final (s)


8 9
Finite element model 4.39  10 1.27  10 39 h, 9 min
8 8
Arbitrary Lagrangian Eulerian model 4.39  10 1.23  10 14 h, 30 min
7 7
Smoothed particle hydrodynamics model 2.07  10 1.98  10 1 h, 4 min
a
CPU-time referred to a Pentium 4–1700 MHz CPU, and 256 MB RAM PC.
ARTICLE IN PRESS

942 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

Fig. 7. Impact of the 21.35 mm radius hailstone. (a) FE; (b) ALE; and (c) SPH models.

impact of a number of hailstones. As a further development of the present work, the feasibility of
the different models for simultaneous impact of a number of hailstones could be considered.
Among the models evaluated, the SPH model of the hailstone proves to be the most suitable to
analyse this phenomenon, though further investigation is recommended.
Concluding, as a result of this research the intake design was modified.
ARTICLE IN PRESS

M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944 943

4. Conclusions

Hailstone impacts represent an actual threat for aircraft structures and, therefore, it is
important to develop numerical models to design hail-proof and high-efficiency structures. In
particular, using LSTC LS-Dyna, three numerical models of hailstone have been developed and
investigated in this research with regard to actual hail impact events: Lagrangian FE, ALE and
SPH model.
Initially, a close correlation with experimental data was obtained for each one of the three
models so that, subsequently, it was possible to analyse the hail impact against an engine nacelle
with a certain confidence. However, before considering the case of interest, in order to verify the
hailstone models, hail impacts against a nacelle nose-lip were simulated using velocities close to
the ones of a documented case, so that it could be possible to make a comparison between
numerical and actual damages. Despite the constructive differences and the lack of further impact
data, the numerical damages were definitively similar to the real dents (in dimensions and
appearance).
Then, simulations using the impact velocity prescribed for the certification of an engine intake
(JAR-E 970 and ACJ E 970) were carried out. As a result, a direct comparison between the three
hail models was possible. In particular, the FE model proves to be suitable to reproduce (only) the
early stages of the impact when the distortion of the mesh is not such to cause inaccuracy, drastic
drops in time-step or premature analysis termination. An erosion criterion could be a solution to
avoid excessive mesh distortion. It was not introduced because, to be effective, it produced
damages far from the evidences collected after real hailstone impacts. On the contrary, the time-
step when using the ALE model remains roughly constant throughout all the simulation.
Unfortunately, the obtained results suffer the typical problems arising from the coupling of
solvers based on different approaches. The SPH model proves to be the most effective. In fact, this
model grants a close numerical-experimental correlation and the required CPU-time is
considerably smaller than the one required by the two other models. Furthermore, the SPH
model effectively reproduces the hailstone behaviour also when the hailstone cracks after impact
and, therefore, it seems the most suitable to analyse the simultaneous impact of a number of
hailstones.
Concluding, it is worth noticing that guided by the results of the analyses performed, the intake
design was modified and certified. This can be regarded as a further confirmation of the usefulness
of an appropriate numerical model as a powerful design tool and as a means to considerably
reduce the number of pre-certification experimental tests.

References

[1] Souter RK, Emerson JB. Summary of available hail literature and the effect of hail on aircraft in flight.
Washington: NACA technical note 2734; 1952. p. 1–33.
[2] Hallquist JO. LS-DYNA theoretical manual, LSTC, May 1998.
[3] Hallquist JO. LS-DYNA keyword user’s manual, LSTC, vol.1–2, March 2001.
[4] Thomson RG, Hayduk RJ. An improved analytical treatment of the denting of thin sheets by hail. NASA
technical note D-6102, Washington, DC, January, 1971, p. 1–36.
ARTICLE IN PRESS

944 M. Anghileri et al. / International Journal of Impact Engineering 31 (2005) 929–944

[5] Thomson RG, Hayduk RJ. An analytical evaluation of the denting of airplane surfaces by hail. NASA technical
note D-5363, Washington, DC, August 1969, p. 1–34.
[6] Belytschko T, Liu WK, Moran B. Non linear finite elements for continua and structures. New York: Wiley; 2000.
[7] Kim H, Kedward KT. Experimental and numerical analysis correlation of hail ice impacting composite structures.
AIAA-99-1366, 1999, p. 1416–26.
[8] Kim H, Kedward KT. Modeling hail ice impacts and predicting impact damage initiation in composite structures.
Am Inst Aeronautics Astronautics 2000;38(7):1278–88.
[9] Schulson EM. The structure and mechanical behavior of ice, Army Research Office. J Met 1999;51(2):21–7.
[10] Anghileri M, Castelletti L-ML, Invernizzi F, Mascheroni M. A numerical model for hail impact analysis. 34th
European rotorcraft forum, September 2004.
[11] Brockman RA, Held TW. Explicit finite element method for transparency impact analysis. University of Dayton
Research Institute, 1991.
[12] Anghileri M, Castelletti L-ML, Lanzi L, Mentuccia F. Composite materials and bird-strike analysis using explicit
finite element commercial codes. Eighth International Conference on Structures Under Shock and Impact (SUSI).
2004.
[13] Accident Investigation Board Finland, Aircraft Damage in Hailstorm West of Helsinki on 21.7.2001. Investigation
report B 5/2001 L , p. 1–31.

You might also like