Professional Documents
Culture Documents
American Transactions on
Engineering & Applied Sciences
http://TuEngr.com/ATEAS
a
Department of Mechanical Engineering, University of Puerto Rico at Mayagüez, PR 00680, USA
b
Engineering Department, Central Connecticut State University, New Britain, CT 06050, USA
ARTICLEINFO A B S T RA C T
Article history: In this second of a three-paper sequence, we developed a
Received December 23, 2012
Received in revised form standard work using the Smoothed Particle Hydrodynamic (SPH)
24 February 2013 approach in LS-DYNA and compared the results against those the
Accepted February 26, 2013 Lagrangian model and available experimental results. First, the SPH
Available online
March 04, 2013 model was validated against a one-dimensional beam centered
Keywords: impact’s analytical solution and the results are within 3% error.
Finite element; Bird-strike events were divided into three separate problems: frontal
Impact analysis; impact on rigid flat plate, 0 and 30 deg impact on deformable tapered
Bird-strike; plate. The bird model was modeled as a cylindrical fluid. We
Smooth-particle successfully identified the most influencing parameters when using
hydrodynamics. SPH in LS-DYNA. The case for 0 deg tapered plate impact shows
little bird-plate interaction because the bird is sliced in two parts and
the results are within 5% difference from the test data available in the
literature, which is an improvement over the Lagrangian model.
Conclusion: The developed SPH approach is suitable for bird-strike
events within 10% error.
1.1 Motivation
As we mentioned in the first paper [1], collisions between a bird and an aircraft are known as
bird-strike events. According to the Federal Aviation Administration (FAA), wildlife strikes cost
the U.S. civil aviation industry over $300 million and more than 500,000 downtime hours each
year [2].
We previously developed the Lagrangian model to predict bird-strike events [1]. However,
this description causes losses of the bird mass due to the fluid behavior of the bird, which causes
large distortions in the bird model. This mass loss may reduce the real loads applied to the fan
blade, which is the reason why the Smoothed Particle Hydrodynamics (SPH) is being studied in
this work. LS-DYNA has the capability to use SPH formulation to model this fluid-structure
interaction problem. In this work, we plan on developing a SPH standard work to modeling
bird-strike events.
Bird-strike events have been studied using Lagrangian method in different finite element
codes [8]. But we seek a model with a better correlation. As we did for the case of Lagrangian
model we will use the work by Moffat et al. [9] and Barber et al. [7] as our reference. The
geometrical model that was used for the bird was a cylinder with spherical ends with an overall
length of 15.24 cm and a diameter of 7.62 cm. The bird density is 950 kg/m3. Moffat et al. [9]
found that the pressures were insensitive to the strength of the bird and a yield stress of 3.45 MPa
Martin [10] studied a transient, material, and geometric nonlinear finite element based impact
analysis using PW/WHAM. His work consisted in simulating soft body impact over stainless steel
disc, a deformable flat plate, and a tapered plate. The formulation employed was very similar to the
concept of the meshless finite element technique SPH. Many authors have presented basic
equations related with the SPH approach. Commonwealth Scientific & Industrial Research
Organization (CSIRO) provided the equations for the SPH approximation using the smoothing
kernel function. Lacome [11] described the conventions used for the selection of the smoothing
length. This is a very important parameter because the spatial resolution of the model depends on
the smoothing length and the characteristic length of the meshed particle. La-come [12] also
provided important information regarding the SPH process, the process of the neighbor search in
A general description of the SPH method was presented by Hut et al. [13]. The authors
presented applications of the method as well as information about the computational parameters for
the SPH method and the expectations for accelerating processing time with the implementation of
faster computers.
In this work, we attempt to create a standard work based on the SPH formulation by
identifying the most important influencing parameters in the bird-strike simulation and validate the
simulation with the test data and compare against the Lagrangian model previously developed
[1].
2. Impact Analysis
The bird-strike events are considered as soft body impact in structural analysis because the
yield point of the bird is far smaller when compared with that of the target. Thus, the bird at the
impact can be considered as a fluid material. The soft body impact results in damage over a larger
area if compared with ballistic impacts. Let us understand the main equations involved in this
study.
(1)
where P represents a diagonal matrix containing only normal pressure components, ρ the density,
and V the velocity vector. The conservation of mass per unit volume equation can be written as
follows:
(2)
(3)
In this formulation, the fluid is represented as a set of moving particles, each one representing
an interpolation point, where all the fluid properties are known. Then, with a regular interpolation
function called smoothing length the solution of the desired quantities can be calculated for all the
particles. A real fluid can be modeled as many fluid particles provided that the particles are small
compared to the scale over which macroscopic properties of the fluid varies, but large enough to
contain many molecules so macroscopic properties can be defined sensibly. A large number of
particles are needed for the SPH calculations, since the continuum limit is recovered when the
*Corresponding author (V. Goyal), Tel.: 1-787-832-4040 Ext. 2111; E-mail:
vijay.goyal@upr.edu. 2013. American Transactions on Engineering & Applied Sciences.
Volume 2 No. 2 ISSN 2229-1652 eISSN 2229-1660 Online Available at 87
http://TuEngr.com/ATEAS/V02/083-107.pdf
number of particles goes to infinity. Particles in the SPH method carry information about their
hydrodynamic and thermodynamic information, this in addition to the mass needed to specify the
evolution of the fluid. Nodes in SPH are similar to nodes in a mesh, the difference is that these
nodes are continuously deformable and distort automatically to put more of the computational
effort in regions of relatively high density.
One disadvantage in SPH is that this method is computationally demanding, both in memory
and in CPU time. This can be overcome using a parallel analysis with more than one CPU. There is
also the difficulty of establishing the boundary condition when using the SPH method. Another
disadvantage is that particles may penetrate the boundaries and causing loss of smoothness and
accuracy.
Figure 1 illustrates an integration cycle in time of the SPH computation process (Lacome
[12]). In the SPH analysis, it is important to know which particle will interact with its neighbors
because the interpolation depends on these interactions. Therefore, a neighboring search technique
has been developed, as shown in Figure 2 (Lacome [12]). The influence of a particle is established
inside of a sphere of radius of 2h, where h is the smoothing length In the neighboring search, it is
also important to list, for each time step, the particles that are inside that sphere. If we have N
particles, then it is required (N − 1) distance comparison. If this comparison is done for each
particle, then the total amount of comparisons will be N(N − 1). It is better to have a variable
smoothing length to avoid problems related with expansion and compression of material. The main
idea of this concept is that it is necessary to keep enough particles in the neighborhood to validate
the approximation of continuum variables. The smoothing is allowed to vary in time and space. For
a constant smoothing length, a material expansion can lead to numerical fracture and a material
For the type of impact problems studied in this work, temperature does not play a big role. In this
context, the term vanishes and thus the equation of state reduces to
The volumetric strain is given by the natural logarithm of the relative volume.
(4)
Where ∆𝑡 is the time it takes to complete the impact. Equation (4) has two unknowns: the average
force and the impact time. The impact time is taken to match the impact time given by LS-DYNA,
and thus perform a fair comparison. Once the impact time in known, the force is obtained straight
forward using Eq. (4).
4.1.1 Bird-Model
How to model a bird is quite challenging and a model that best meets the testing bird properties
must be used. The model considered in this work is that of a cylinder, as suggested by Bowman and
Frank [15]. It has been shown that this model of bird produce close results to real birds. In
To model the bird, first a square mesh with equivalent sides to the pre-calculated diameter are
constructed, and then extruded to a length with different number of subdivisions, as seen in Figure
9(a). The elements located in the square mesh are eliminated to approximate the mesh shape to
that of the circular (cylindrical) mesh used in the Lagrangian simulation, as shown in Figs. 9(b) and
9(c). The remaining nodes are used to calculate the mass and dimensions of each spherical
element, and the SPH particles were constructed in each remaining node (Goyal et al. [16]). The
lumped mass for each SPH element is calculated using the following equation:
(5)
Note that the number of nodes is equal to the number SPH elements. For the shot reproduced the
dimensions of the bird (cylindrical representation) were: D = 41.4431 mm, L = 60 mm, ρ = 912.60
kg/m3, V = 80936.7 mm3, and m = 0.073863 kg. Table 1 summarizes the main properties used for
modeling the bird.
Figure 10: Graphical representation of Pressure Time oscillograph for shot 4992-B [7].
Here we compared the tendencies on the pressure-load curves of the computer simulations
with experimental data, and determined velocities and densities based on the mass and velocity
quantities of the test data. To have a fair comparison of the results obtained by the bird-strike
simulation using LS-DYNA and the Barber et al. [7] research it is necessary to reproduce as
accurate as possible the results displayed from the pressure transducers recording the impact event.
Because of not having the digital experimental data of the test results to be reproduced, the graphs
presented on Barber’s report are imported into MatLab to obtain approximate readings, which are
later compared with those obtained by LS-DYNA.
Figure 11: Geometric model for the Lagrangian bird and Target shell.
The first simulation consisted in the bird strike event against solid rigid plates as done by
Barber et al. [7]. The second simulation model was based in the work of Moffat et al. [9],
bird-strikes on tapered plates. All simulations using the SPH model are compared with their
respective available test data and the Lagrangian models (Goyal et al. [1]). Now we proceed to
explain the results obtained by the SPH model using LS-DYNA.
Here, we explain the variables that affect directly the output data for a bird-strike event
modeled in LS-DYNA using the SPH approach. The two most important parameters that
influenced the desired response are the type of contact between the SPH particles and the target
surface, and the Particle Approximation Theory (keyword: FORM inside the SPH card
*CONTROL_SPH). The contacts used for the simulations performed included only those that
Figure 13: Deformation of the SPH bird for different time intervals.
The FORM parameter (Particle Approximation Theory parameter) inside the *CONTROL
SPH card can be changed and set to FORM=1 (remoralization approximation). This generates a
deformation similar to that predicted by Wilbeck and Barber [7] of the impacting cylinder reducing
the effect of bouncing of the particles when entering in contact with the target. By changing this
parameter, a bounce-off effect is produced and thus it affects the interaction of the upcoming
The variable NCBS (Number of Cycles Between particle Sorting) of the *CONTROL SPH
card is varied between 2 and 5. The same force at the interface surface is used for each value for the
NCBS variable. Figure 13 shows the deformation at different time intervals for the case when
NCBS is 2 and 5. The final deformation is in good agreement to that predicted by Wilbeck and
Barber [7]. There is no difference between the resultant force at the interface of the SPH bird and
the Shell Target and have a peak value of 5.4 kN. Thus, NCBS is a nonaffecting variable in the
simulation. The contact used for this simulation was *CONTACT CONSTRAINT NODE TO
SURFACE contact type.
Figure 14: Deformation of the SPH bird at different time intervals using a transducer.
Table 3: Peak impact Pressure comparison between SPH and Lagrangian simulations.
Figure 16: SPH simulations of shot 5126 A for different number of particles.
Figure 16 shows different deformation plots for SPH simulations using different number of
particles in the model. As explained, this variation in number is considered in order to study its
influence in the stability of the model, in the deformation and in the resultant force plot along the
whole simulation. The simulation obtained with 2464 particles was unstable, which could be
caused by the non-uniform distribution of the particles across the entire model of the bird. All the
other models were constructed with a more uniform distribution of the particles and with a distance
between adjacent particles almost constant. These models produced deformations close to
experimental data. Models with 4700, 8700 and 22,000 SPH particles showed deformation with a
sliding behavior in the steady phase of the impact. In contrast, models using 5000 and 10000 SPH
particles presented bouncing of the particles in the steady phase of the bird-strike. However this
bouncing effect is more affected by the type of contact used than for the amount of SPH particles
100 Vijay K. Goyal, Carlos A. Huertas, and Thomas J. Vasko
used for the creation of the model, as shown in Figure 17.
Figure 18: Comparison of peak pressure obtained in Lagrangian and SPH simulation using the
nodes to surface contact TEROD=1.1.
Using the results obtained it is observed that the mesh density affects the simulation.
Simulated transducers make possible getting the pressure distribution graph. When two different
mesh resolutions are used, the results are more accurate with respect to the maximum pressure
obtained.
Figure 19: Result of the interaction for the SPH simulation of the tapered plate impact using
material NULL.
Table 4: Peak force comparison for various SPH tapered plate impact at 0 degrees.
When using elastic fluid material, there was interaction of the plate and the bird, but as the
impact is a 0 degrees the plate slices the bird in two parts. This occurs because the bird impacts at
the narrowest side of the plate. Figure 19 shows that the maximum peak force computed for this
case was 0.001694 MN. This force is 88.66% lower than the computed for the Lagrangian case.
One reason for this is that the elastic fluid accounts for the deviatoric stresses, computing the
stresses that cause deformation in the solid elements of the bird. The impact energy therefore will
be absorbed for the material of the bird, decreasing the force generated in the impact. Table 4 lists
the peak forces obtained for the Lagrange and SPH simulations of the bird-strike on tapered plates
impacting at an angle of 0 degrees. The simulation that best converges to the Lagrangian case is
the SPH simulation using the material null for a SPH model with 26,000 particles.
Figure 20: Result of the interaction for the SPH simulation of the tapered plate.
Figure 21: Deformation of the tapered plate using material null using SPH model.
Table 5: Maximum normal deflection for a bird impacting a tapered plate at 30°.
Other cases where also ran and results are enclosed in Table 5. Thus, the simulation that has
better convergence to the maximum deflection found by Moffat is the SPH simulation using
material null and 26,000 particles. In addition, for this model the peak force obtained was only
13.4% different from the one calculated using the Lagrange simulation with material elastic fluid.
But if the peak force is compared with the Lagrangian case using the same null material the error is
8.93%. However, changing the material in the SPH simulation to the material elastic fluid does not
result in a convergence of the force to the Lagrangian model using the same elastic fluid material.
For the first case, the frontal impact against a flat rigid plate, a bird model with 8700 particles
with one-way constraint contact type and a null material type can be used. The error obtained in the
pressure after comparing with the experimental data is within 10%. For the Lagrangian model the
error when using elastic fluid material and an eroding contact type produced and error of 9%.
Therefore, the SPH model can successfully be used in this kind of frontal impacts.
For the second case, 0 and 30 degrees impact against a tapered plate, the SPH model validates
the results obtained in the literature. The SPH model is recommended over the Lagrangian model
for angled impacts on tapered plates since the error of the maximum normal deflection in less than
7% when compared with the experimental data available in literature. These results are validated
using 26,000 SPH particles, a eroding contact type and a null material. For the 0 degree impact,
neither the Lagrangian nor the SPH provide any significant deformation. Results show very little
interaction between the bird and the tapered plate because the bird is being sliced in two parts,
being the null material the best approximation when using SPH.
Based on the results, the SPH method can be considered as a good alternative to simulate
bird-strike events although it is more complex in the model creation.
7. Acknowledgments
This work was performed under the grant number 24108 from the United Technologies Co.,
8. References
[1] V. K. Goyal, C. A. Huertas, T. J. Vasko, 2013. Bird-Strike Modeling Based on the Lagrangian
Formulation Using LS-DYNA. Am. Trans. Eng. Appl. Sci. 2(2): 057-081. Available at:
http://TuEngr.com/ATEAS/V02/057-081.pdf. Accessed: March 2013.
[2] J. Metrisin, B. Potter, Simulating Bird Strike Damage in Jet Engines, ANSYS Solutions 3 (4)
(2001) 8–9.
[3] E. Parkes, “The permanent deformation of a cantilever strucks transversely at its tip”, in:
Proceedings Roy. Soc. Lond., England, 1995.
[4] W. J. Stronge, T. Yu, Dynamic Models for Structural Plasticity, Springer-Verlag, London,
Great Britain, 1993.
[5] W. Goldsmith, IMPACT: The Theory and Phisical Behaviour of Colliding Solids, Dover
Publications, Mineola, New York, 2001.
[9] W. Moffat, Timothy J. and Cleghorn, “Prediction of Bird Impact Pressures and Damage Using
MSC/DYTRAN”, in: Proceedings of ASME TURBOEXPO, Louisiana, 2001.
[10] N. F. Martin, Nonlinear Finite Element Analysis to Predict Fan Blade Impact Damage.
[11] J. Lacome, “Smooth Particle Hydrodynamics (SPH): A New Feature in LS-DYNA”, in:
Proceedings of the 6th International LS-DYNA Users Conference, 2000.
[15] D. R. Bowman, G. J. Frank, “IBRG ARTIFICIAL BIRD PROJECT”, Work programmed and
Schedule, United Kingdom (2000).
Dr. V. Goyal is an associate professor committed to develop a strong sponsored research program for
aerospace, automotive, biomechanical and naval structures by advancing modern computational
methods and creating new ones, establishing state-of-the-art testing laboratories, and teaching
courses for undergraduate and graduate programs. Dr. Goyal, US citizen and fully bilingual in both
English and Spanish, has over 17 years of experience in advanced computational methods applied to
structures. He has over 15 technical publications with another three in the pipeline, author of two
books (Aircraft Structures for Engineers and Finite Element Analysis) and has been recipient of
several research grants from Lockheed Martin Co., ONR, and Pratt & Whitney.
C. Huertas completed his master’s degree at University of Puerto Rico at Mayagüez in 2006.
Currently, his is back to his home town in Peru working as an engineer.
Dr. Thomas J. Vasko, Assistant Professor, joined the Department of Engineering at Central
Connecticut State University in the fall 2008 semester after 31 years with United Technologies
Corporation (UTC), where he was a Pratt & Whitney Fellow in Computational Structural
Mechanics. While at UTC, Vasko held adjunct instructor faculty positions at the University of
Hartford and RPI Groton. He holds a Ph.D. in M.E. from the University of Connecticut, an M.S.M.E.
from RPI, and a B.S.M.E. from Lehigh University. He is a licensed Professional Engineer in
Connecticut and he is on the Board of Directors of the Connecticut Society of Professional Engineers
Peer Review: This article has been internationally peer-reviewed and accepted for
publication according to the guidelines given at the journal’s website.