ARTICLE IN PRESS

International Journal of Impact Engineering 32 (2006) 928–946

www.elsevier.com/locate/ijimpeng

Scaling numerical models for hypervelocity test sled

slipper-rail impacts
A.G. Szmerekovsky, A.N. Palazotto, W.P. Baker
Department of Aerospace Engineering, Air Force Institute of Technology, 2950 Hobson Way, Wright-Patterson AFB,
OH 45433-7765, USA
Received 9 March 2004; received in revised form 20 July 2004; accepted 5 September 2004
Available online 26 November 2004

Abstract

Hypervelocity test sled slipper-rail impacts have been simulated numerically using the finite volume
hydrocode, ChartD to the Three-Halves (CTH). This study addresses the difficulties of applying CTH
model solutions to real test sled runs. Past CTH models using dimensions different than actual test sleds
have been used to study phenomenological aspects of the problem. However, quantitative results from the
CTH model solution do not apply directly to actual test sled runs due to strain rate effects and time scale
differences. The Buckingham Pi Theorem is applied to two potential hypervelocity gouging models.
Validity of the invariant products is tested using sample CTH hypervelocity gouging models that are scaled
up to simulate dimensions of a real test sled. Real test sled dimensions are desired in order to more closely
simulate actual test sled runs. Invariant products developed from application of the Buckingham Pi
Theorem can be used as guidelines for determining whether a CTH model is applicable to a test sled with
specific dimensions. Strain rate effects are investigated to study whether deviations between scaled CTH
models may be reduced by modifying the constitutive model.
Published by Elsevier Ltd.

Keywords: CTH; Dimensional analysis; Gouging; Hypervelocity impact; Scaling

Corresponding author. Tel.: +1 937 255 3636x4599; fax: +1 937 656 7621.
E-mail addresses: andrew.szmerekovsky@afit.edu (A.G. Szmerekovsky), anthony.palazotto@afit.edu
(A.N. Palazotto), william.baker@afit.edu (W.P. Baker).

0734-743X/$ - see front matter Published by Elsevier Ltd.

doi:10.1016/j.ijimpeng.2004.09.011
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 929

1. Introduction

Gouging continues to be a concern for hypervelocity rail test sleds. The effects of gouging vary
from requiring repair to the rail to catastrophic failure. If one observes Fig. 1, which shows the
important divisions of the Holloman High Speed Test Track (HHSTT) rocket sled, it can be
stated that the rocket is held to the rail using slippers (also called shoes). These slippers are made
of VascoMax 300, a high strength maraging steel. The slippers are not tight, leaving a small gap
on one side or another. This results in the sled riding in free flight with intermittent contact on the
steel rail (1080 steel). The impact between the slipper and rail may result in gouging. The main
divisions of the rocket sled that are discussed in this paper are the sled system (i.e., the rocket,
payload, and other structures), the rails, and the slippers.
Gouges are characterized by the shallow removal of material from the rail and the slipper
(as shown in Fig. 2) and have been observed to occur at sled speeds greater than 1.5 km/s.
Additionally, the problem is quite complex in which materials may change phase, stress waves
become important, and heat generation must be traced. The current means of mitigating gouges is
to coat the rail with a polymer, such as epoxy. This has reduced the occurrence of gouging, but it
is not possible to obtain a uniform coating thickness with current application methods. As
velocities increase, it is not understood whether coating variations will act to increase gouging or

Fig. 1. HHSTT Rocket Sled with sled system, slippers, and rails labeled.

Fig. 2. Actual gouge at a reentrant corner of the rail in which a plane-strain condition likely exists at impact.
 ARTICLE IN PRESS

930 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

continue to mitigate the phenomenon during test sled runs. In addition, the effect of
nonequilibrium thermodynamics on the thermal environment prior to impact of the slipper/
shoe with the rail/guider has not been investigated fully. Each scenario listed must be studied, but
it would do no good if the model did not depict real dimensionality.
Gouging has been studied using test track runs and observations [1–6] and experimentally using
rail guns [7–9]. These findings indicate that gouging occurs under conditions of high-speed sliding,
high temperature, large strain rate, plasticity, and high stress. They also suggest there is a mixing
of slipper and rail materials during gouging.
Gouging has also been studied numerically using the finite-volume hydrocode CTH [10]. Direct
application of these numerical results to actual hypervelocity gouging requires subjective
engineering judgment. For instance, slipper models with vertical impact velocities designed to
match the kinetic energy of test sled vertical impacts have been used [11–13] as one means of
approximating a real impact. However, this approach does not approximate the momentum of the
impact unless the test sled mass is also simulated. As a consequence, results based on this model
can only be applied from a phenomenological perspective.
Previous CTH models were reviewed. These included both oblique impact and asperity initiator
models. Difficulties arise when one tries to apply the results from these phenomenological CTH
models to a real test sled. There are discrepancies between dimensions in the CTH models and real
test sled dimensions. Previous CTH models either contain dimensions that are orders of
magnitude smaller (Barker [14,15] and Schmitz et al.’s [16] asperity models) than a real slipper or
have one or more dimensions that are significantly different (Tachau’s [17], Tachau et al.’s [18],
and Laird’s [19–22] models). One consequence of this difference in dimensions is that strain rate
effects in the smaller CTH model are much higher than a model with real dimensions. Another is
the different event times between the CTH solution and a comparative model with real
dimensions. A means of determining if results from previous CTH gouging models can be applied
to a given test sled, is desired. This will also provide insight into developing a new CTH model
that has direct application to a real test sled run.
Similarity methods have been applied to structural impacts [23,24] and more specifically to
hypervelocity impacts [25–28]. These methods may be applied to the slipper-rail impacts of a
hypervelocity test sled. It has been shown that geometric similarity of ballistic impacts does not
properly scale material strain rate sensitivity, thermal conductivity, and fracture.
This study applies the Buckingham Pi Theorem to two potential hypervelocity gouging models.
Using invariant parameters from the dimensional analysis, a sample CTH hypervelocity gouging
model is scaled up to better simulate the dimensions of a real test sled. The results between the
base model and the scaled-up version are compared to determine if satisfying the theoretical
Buckingham Pi invariants results in similar numerical solutions. Strain rate parameters are
explored to determine their effect on the solutions of scaled numerical models.
If the invariant products are found to be an accurate means of scaling CTH models, they may
be used as a guide for qualitative application of CTH numerical results to actual test sleds. For
example, invariant products derived from real test sled dimensions can be compared to the same
invariant products of a CTH numerical model. If the invariant products are the same, the results
from the CTH model may apply to the actual test sled, depending on strain rate sensitivity. If they
do not match, then the invariant products guide changes to the CTH model so the solution can
apply to the given test sled.
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 931

The computational algorithm in CTH uses a finite volume method to simultaneously solve the
nonlinear and time-dependent system of equations of conservation, equation of state, and
constitutive model equations for problems of high-energy impact such as the hypervelocity
gouging problem. But not every dimensioned quantity in the problem is under direct control of
the user (e.g., when using tabular equations of state). Satisfying the invariant products does not
necessarily scale these other quantities properly.
Also, dimensional analysis does not indicate the relative significance of specific invariant
products. For example, an invariant product that varies by approximately 10 percent between
models, could result in a solution that is 50 percent different. Nonlinear relationships between
various solution parameters means that great care must be used when drawing conclusions about
real test sleds, based on CTH models. Thus, an investigation into the effect of not properly scaling
quantities such as strain rate, is important.
Finally, there are additional considerations—such as phase changes, shock, and fracture—that
provide potential bifurcations in which small variation in one parameter results in conditions that
require a totally different approach in the scaled model. For example, variation of one parameter
may result in damage for a scaled model. This damage results in a solution that neither
quantitatively nor qualitatively matches any trend in the invariant products.
In this paper, the Buckingham Pi Theorem is applied to oblique (both horizontal and vertical velocity
components) hypervelocity impacts of a slipper on a clean prismatic rail and on a coated misaligned
rail. The rail misalignment is modeled as a semi-elliptic asperity on the flat surface of the rail.
Dimensional analysis of the hypervelocity gouging problem in this study uses the conservation
equations as the physical laws that depict the controlling relationship. It is initially assumed for
purposes of the dimensional analysis that the slipper and rail have the same or similar material
properties. Therefore, original density, ro is assumed to hold the same value as both r for the
slipper and r for the rail. Likewise, original shear modulus, G o holds the same value as both G for
the slipper and G for the rail.

2. Dimensional analysis

In order to conserve space in this paper, this dimensional analysis considers in detail only two
different cases that can be found in a real test sled hypervelocity run. The first case is a slipper-rail
impact in which the velocity vector is at an extremely shallow angle upon a clean, flat rail. Models
1 and 2 are used to depict this case. Model 1 is used as a sample CTH legacy model that has been
used to depict hypervelocity gouging using dimensions smaller than an actual test sled. Model 2 is
used to simulate actual test sled dimensions. Dimensions of model 1 are scaled up by a factor of
two to create model 2.
The second case is a slipper-rail impact in which the velocity vector is at an extremely shallow
angle upon a rail coated with a polymer material such as epoxy. Gouging is initiated when the
coating is penetrated and the slipper impacts the rail-misalignment. A schematic of this case is
shown in Fig. 3. Models 3 and 4 are used to depict this case. Models 1 and 2 are really just
simplified versions of this case. All geometric dimensions of model 3 are half that of model 4.
Model 4 simulates the real test sled dimensions and model 3 simulates a sample CTH model.
The fundamental units derived from the dimensioned quantities are mass, M; length, L; and
time, T. The dimensioned quantities and their corresponding fundamental units are shown in
 ARTICLE IN PRESS

932 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 3. Simplified model of an obliquely impacting slipper on a rail roughness with a coating for dimensional analysis.

Table 1
Dimensioned quantities and their corresponding fundamental units

Dimensioned Parameter Fundamental

quantity units

m Initial bulk sled mass M

ro Original density of the slipper M L
3
roc Original density of the coating M L
3
‘ Slipper length L
h Slipper height L
w Slipper width L
tc Coating thickness L
ha Asperity/rail roughness height L
Aa Asperity/rail roughness area L2
ux Horizontal velocity L T
1
uy Vertical velocity L T
1
c Slipper material speed of sound L T
1
cc Coating material speed of sound L T
1
E Initial total energy of the bulk sled system M L2 T
2
S Constant energy source of the bulk sled system M L2 T
2
sy;c Slipper critical yield strength M L
1 T
2
sy;cc Coating critical yield strength M L
1 T
2
Eo Original slipper elastic modulus M L
1 T
2
E oc Original coating elastic modulus M L
1 T
2
Go Original slipper shear modulus M L
1 T
2
G oc Original coating shear modulus M L
1 T
2

Table 1. They were selected by assuming the slipper and rail material properties having the same
values. The independent variables for the model are horizontal position, x; vertical position, y;
lateral position, z; and time, t: Scaled coordinates and time values are used when making
comparisons between models within each case.
Table 2 lists the dimensioned quantities and their values for each of the models used in the
numerical study. Table 3 lists the material properties used for the slipper and rail. Not listed are
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 933

Table 2
Computational models used to investigate scaling laws

Dimensioned Parameter values

quantity Model 1 Model 2 Model 3 Model 4

m 56.75 kg 454 kg 56.75 kg 454 kg

ro 8:129 g cm
3 8:129 g cm
3 8:129 g cm
3 8:129 g cm
3
roc — — 1:186 g cm
3 1:186 g cm
3
‘ 4.37 cm 8.74 cm 4.37 cm 8.74 cm
h 2.54 cm 5.08 cm 2.54 cm 5.08 cm
w 10.80 cm 21.60 cm 10.80 cm 21.60 cm
tc — — 0.00762 cm 0.01524 cm
ha — — 0.03 cm 0.06 cm
Aa — — 0:1296 cm2 0:5184 cm2
ux 2 km s
1 2 km s
1 2 km s
1 2 km s
1
uy 50 m s
1 50 m s
1 50 m s
1 50 m s
1
c 398; 000 cm s
1 398; 000 cm s
1 398; 000 cm s
1 398; 000 cm s
1
cc — — 273; 000 cm s
1 273; 000 cm s
1
E 113.6 MN 908.6 MN 113.6 MN 908.6 MN
S — — — —
sy;c 1447 MPa 1447 MPa 1447 MPa 1447 MPa
sy;cc — — 15 MPa 15 MPa
Eo 184.2 GPa 184.2 GPa 184.2 GPa 184.2 GPa
E oc — — 1.10 GPa 1.10 GPa
Go 71.8 GPa 71.8 GPa 71.8 GPa 71.8 GPa
G oc — — 0.4 GPa 0.4 GPa

Table 3
Material model constants

Johnson–Cook 1080 Steel (Iron) Rail Steinberg–Guinan–Lund VascoMax 300a Slipper

Model constant Value Units Model constant Value Units

ro b 7.850 g cm
3 ro 8.129 g cm
3

A 1:7526  109 g cm
1 s2 A 2:06  10
12 g cm
1 s2
B 3:8019  109 g cm
1 s2 Go 7:18  1011 g cm
1 s2
Go b 7:8  1011 g cm
1 s2 Yo 1:447  1010 g cm
1 s2
Y ob 7  109 g cm
1 s2 Y max 2:5  1010 g cm
1 s2
TM 1835.7 K T mo 2310 K
C 0.06 — B 3:15  10
4 1 K
1
n 0.32 — n 0.5 —
m 0.55 — b 2.0 —
go 1.67 —
a 1.2 —
a
Constants not listed are zero.
b
Not used in the Johnson–Cook model, listed for reference.
 ARTICLE IN PRESS

934 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

the material models for the epoxy coating which is a Von Mises elastic–plastic constitutive model
with a Mie–Grüneisen equation of state. Table 4 lists the similitude invariants determined in the
dimensional analysis and the values for each model based on the dimensioned quantities from
Table 2.

3. Numerical investigation

Invariants remain constant between scaled models so that the solutions are comparable. We
compare results for model 1 with model 2 and results from model 3 with model 4. An arbitrary
mass of 56.75 kg is selected for models 1 and 3 in order to add momentum as well as kinetic energy
to the system. This material is different from the slipper material for the following reasons:
(1) The actual sled system consists of materials that are different than the slipper.
(2) It is assumed that gouging occurs locally in the slipper material. This means the sled system
mass only affects the gouging phenomena by adding to the effective mass of the impact which
affects the total energy and momentum of the impact. Assuming this to be true, the material
properties of the sled system mass may not be critical to studying the gouging phenomenon
unless the phenomenon occurs after shock reflection off the upper surface of the slipper. For
the cases under study, shock reflection at the interface between slipper and artificial mass
materials will occur at approximately 6:38 ms in models 1 and 3. The solutions for the time
periods under consideration do not contain shock reflections.

Platinum is chosen for the artificial mass material in this study because of its high density,
which minimizes the dimensions of the artificial mass and reduces the model size, thus reducing
the number of cells for analysis.
Models 1 and 3 are compared to their scaled-up counterparts in models 2 and 4, respectively.
The material properties remain the same between models and the horizontal and vertical impact
velocities also remain the same. The invariant products must remain constant between models for
the results to be consistent.
The time scale used for the initial models are T~ 1 ¼ ‘=ux : The time scale for the scaled-up models
is T~ 2 ¼ 2 ‘=ux : This time scale provides a means for comparing results between models
(i.e., 2T~ 1 ¼ T~ 2 ).
The solution is affected by material mixing in the CTH algorithm. During the study, it was
found that material boundaries between sliding materials must coincide with Eulerian mesh cell
boundaries. This is important for avoiding problems due to numerical instabilities of the sliding
interface algorithm.
Model 1 is shown in Fig. 4. A tracer point in the slipper of models 1 and 2 is first selected close
to the surface of interaction (one cell length away) to capture the effect of high-strain rates for the
impact between the slipper and rail. This traces the worst case scenario since this point becomes
part of the gouged material, which undergoes high-strain rates.
The solutions for pressure and deviatoric stress are plotted in terms of model 2 time.
A coarse and a refined mesh were both studied. The refined mesh results are shown here.
Mesh refinement was seen to be a factor in convergence of the solution. The plots of results from
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 935

Table 4
Invariants and their values for numerical models

Invariants Model 1 Model 2 Model 3 Model 4

Mass scaling
‘3 0.0120 0.0120 0.0120 0.0120
p1 ¼ ro
m
‘3 — — 0.00174 0.00174
p2 ¼ roc
m

Geometric scaling
h 0.5812 0.5812 0.5812 0.5812
p3 ¼
‘
w 2.4714 2.4714 2.4714 2.4714
p4 ¼
‘
tc — — 0.00174 0.00174
p5 ¼
‘
ha — — 0.00686 0.00686
p6 ¼
‘
Aa — — 0.00678 0.00678
p7 ¼ 2
‘

Velocity scaling
uy 0.025 0.025 0.025 0.025
p8 ¼
ux
c 1.990 1.990 1.990 1.990
p9 ¼
ux
cc — — 1.365 1.365
p10 ¼
ux

Energy scaling
E 0.5 0.5 0.5 0.5
p11 ¼
mu2x
S — — — —
p12 ¼
mu2x
‘3 0.00053 0.00053 0.00053 0.00053
p13 ¼ sy;c 2
mux
‘3 — — 5:5  10
7 5:5  10
7
p14 ¼ sy;cc
mu2x
‘3 0.0677 0.0677 0.0677 0.0677
p15 ¼ E o 2
mux
‘3 — — 0.0004 0.0004
p16 ¼ E oc
mu2x
‘3 0.0264 0.0264 0.0264 0.0264
p17 ¼ G o 2
mux
‘3 — — 0.00014 0.00014
p18 ¼ G oc
mu2x

Subscript a indicates property of asperity/rail roughness.

Subscript c indicates property of the coating material.
 ARTICLE IN PRESS

936 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 4. Model 1 with artificial mass and zoomed in view of tracer placement.

Fig. 5. Time-scaled history comparison of pressure for slipper tracer near oblique impact surface with mesh refinement.

models 1 and 2 are shown in Figs. 5 and 6. The results for the coarser mesh resulted in relatively
larger differences in the solutions for a point near the impact surface than for the refined mesh.
The average difference in pressure plots between models 1 and 2 for the coarse mesh configuration
is around 1.6 GPa over the time period of 7:0 ms: The average difference in deviatoric stress
is 323 MPa.
With improved mesh resolution, the solutions matched more closely. The mesh cell size was
reduced by half, thus increasing the number of cells by a squared term in the vicinity of the tracers.
This was the smallest mesh cell size possible with the computer resources available. The average
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 937

Fig. 6. Time-scaled history comparison of XY deviatoric stress for slipper tracer near oblique impact surface with mesh
refinement.

difference between model results for the improved mesh resolution decreases to 547 MPa with a
maximum difference of 1.5 GPa for pressure, and 93 MPa with a maximum difference of 400 MPa
in deviatoric stress. This finer mesh is used for the next case studied.
The important differences between geometrically scaled models depends on the
phenomenon being studied and the material properties of the materials. Plasticity is an important
factor in development of hypervelocity gouging in a slipper and rail impact. However,
global parameters such as development time of the gouge and size of the gouge are perhaps
of greater interest than equivalency of the stress tensor at a particular point in scaled space
and time.
The next case is shown in Fig. 7. The figure shows where tracers in model 4 are located. Pressure
and deviatoric stress traces are compared for models 3 and 4 in terms of model 4 scaled time. This
case contains a coated rail with rail misalignment, also known as rail roughness. Tracers at
various locations in the slipper, rail, and coating are studied. The tracer locations were selected so
the effects of high-strain rate conditions could be compared to areas of relatively low-strain rate
and stress. The Lagrangian tracer located in the slipper is 200 cells away from the impact surface.
These plots are shown in Figs. 8 and 9. The largest difference between the solutions in pressure is
30 MPa. The largest stress deviator difference is only 22 MPa. Note however, the value of the
stress impulse for the impact at this location is much smaller than near the surface for the
uncoated case of models 1 and 2. Strain rates are much lower here.
The Lagrangian tracer located in the rail in models 3 and 4 is 10 cells away from the impact
surface. These plots are shown in Figs. 10 and 11. The pressure wave is about four times greater at
this point than at the slipper tracer and occurs earlier. The largest difference between the solutions
is 70 MPa for pressure. However, there is a greater increase in difference between models in stress
deviator behavior. The pressure pulse occurs at 2:5 ms in terms of model 4 time. At this time, the
 ARTICLE IN PRESS

938 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 7. Model 4 with artificial mass, coated rail, and rail roughness with zoomed in view of tracer placement.

Fig. 8. Time-scaled history comparison of pressure for slipper tracer in coated models with rail roughness.

largest difference in deviatoric stress of 63 MPa occurs. In terms of absolute value, this number is
on the same scale as the difference in the pressure plot. Strain rate effects show up in this vicinity
of the impact.
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 939

Fig. 9. Time-scaled history comparison of XY deviatoric stress for slipper tracer in coated models with rail roughness.

Fig. 10. Time-scaled history comparison of pressure for rail tracer in coated models with rail roughness.

The Lagrangian tracers located within the coating are near the point of impact. These plots are
shown in Figs. 12–15. The first coating tracer is located directly above the rail tracer and the
second coating tracer is located within the coating on the rail misalignment. The pressure wave,
 ARTICLE IN PRESS

940 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 11. Time-scaled history comparison of XY deviatoric stress for rail tracer in coated models with rail roughness.

Fig. 12. Time-scaled history comparison of pressure for first coating tracer in coated models with rail roughness.

depicted in Fig. 12, is almost 10 times greater at this point than at the slipper tracer. The average
difference between pressure solutions for the first coating tracer is 19 MPa. Compare this to the
67 MPa average deviation of pressure over 5:0 ms for the second coating tracer (in which a larger
pressure pulse occurs). Deviatoric stresses in the coating remain relatively low at the first coating
tracer where the average difference between models is only 1.13 MPa. There is a marked increase
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 941

Fig. 13. Time-scaled history comparison of XY deviatoric stress for first coating tracer in coated models with rail
roughness.

Fig. 14. Time-scaled history comparison of pressure for second coating tracer in coated models with rail roughness.

in average difference between models at the second coating tracer (up to 54 MPa), even
accounting for the greater deviatoric stresses in the second coating tracer. The pressure pulse in
Fig. 14 occurs at 3:5 ms in terms of the time of model 4. After this large pressure, strain rate effects
lead to a maximum difference of 200 MPa in the stress deviator for the second coating tracer.
 ARTICLE IN PRESS

942 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 15. Time-scaled history comparison of XY deviatoric stress for second coating tracer in coated models with rail
roughness.

Time histories show important differences on a local scale. Globally, the results match well
between models. Deviations between geometrically similar models are shown to increase with time
and after high-pressure waves where large strain rate effects are dominant. These results are
summed up in terms of average percentage difference in Table 5.
One can see that there are so many relations inside CTH that would have to be adjusted, that it
becomes near impossible to incorporate the Buckingham Pi parameters for each of these. One
may trace the solution of the two step method considering the Lagrangian and remapped Eulerian
steps to see the possible parameters that could be affected. McGlaun et al. [29] sets out many of
these parameters in their discussion of CTH. We considered only one relation, the Johnson and
Cook constitutive equation, and found a significant change in comparisons. Other parameters
could be associated with the equation of state, thermal softening and many of the thermodynamic
routines, to name a few.
In an attempt to reduce the difference in results between geometrically scaled models due to
material strain rate sensitivity, changes to the constitutive model are investigated. Plastic behavior
of the rail material is characterized in part by the Johnson–Cook constitutive equation:
Y ¼ ½A þ Bð
p ÞN ½1 þ C lnðmaxð0:002;
_p ÞÞ½1
ymh ; (1)
where
T
Tr
yh ¼ : (2)
TM
Tr
The reference temperature, T r is room temperature. A; B; C; N; and m are constants that are
material dependent,
p is plastic strain, and
_p is plastic strain rate. C helps determine the
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 943

Table 5
Summary of difference between models in terms of approximate percent deviation from baseline

Tracer location Ave. % diff. Ave. % diff. Ave. % diff. Ave. % diff.
press. dev. stress press. dev. stress

Models 1 and 2 comparison

Refined mesh
Time interval p3 ms 43 ms

Slipper 5 o5 15 35

Models 3 and 4 comparisons

Time interval p2 ms 42 ms

Slipper o5 o5 o5 o5
Rail o5 5 o5 35
Coating 1 o5 5 o5 17
Coating 2 0a 5 5 100
a
No pressure changes at that location until after 2 ms:

sensitivity of the material to the strain rate. Strain rates in geometrically scaled models cannot be
scaled properly. The strain rate is always larger in the smaller model of a geometrically scaled pair
[23]. In this investigation, the Johnson–Cook model constant C was reduced by a factor of about
10 percent for the CTH model (i.e., model 3) in an attempt to modify strain rate effects and
produce a closer match in scaled results in deviatoric stress with the ‘‘real’’ sled dimensions. The
Johnson–Cook constitutive model is used for the rail material. Data for the rail tracer in the
comparison between models 3 and 4 is shown. In order to scale the constitutive equation
consistently, constants A and B of the model are increased by the same factor that C is decreased
by. This is to keep the overall change in dynamic flow stress consistent as much as possible.
However, some inconsistencies remain. Specifically, a factor of 1/1.1 is used to adjust the strain
rate:
 
p N 1
Y ¼ 1:1½A þ Bð
Þ  1 þ C lnðmaxð0:002;
_ ÞÞ ½1
ymh ;
p
(3)
1:1

where the constant 1 in the second term with C should be 1/1.1 to maintain consistency in the
model.
Since this value cannot be changed, we are effectively changing that term to 1.1. This will have
an overall effect of increasing the dynamic yield strength while decreasing the effect of strain rate
on the constitutive model.
The results of changing the Johnson–Cook constants for deviatoric stress are shown in Fig. 16.
The constants A and B have the fundamental units M L
1 T
2 : These then become pi invariants
that cause adjustments to the existing pi invariants. A and B must be scaled with the kinetic energy
term, ‘3 =mu2x : Changing A and B to 1:1A and 1:1B while maintaining geometric scaling means that
ux must change to 1:05ux :
 ARTICLE IN PRESS

944 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

Fig. 16. Time-scaled history comparison of XY deviatoric stress after modifying Johnson–Cook constants.

This in turn, requires changes to other parameters. The time scale changes to 1:9T~ 1 ¼ T~ 2 where
T 1 is a given time in the modified model 3 and T~ 2 is the equivalent time in model 4. The average
~
difference between deviatoric stress in the models after making these changes remains
approximately the same, but the trends of the stress deviator in model 3 more closely match
the solution using the ‘‘real’’ dimensions of model 4 after adjusting model 3’s constitutive model
for strain rate effects. Adjusting for time, the average difference decreases from around 35% to
about 25%. This is done without significant changes to the pressure.

4. Conclusions

Scaling of numerical models of hypervelocity test sled slipper-rail impacts for CTH has been
presented. The study showed invariant products that may be used to guide application of previous
CTH hypervelocity gouging models to real test sleds. The means to do this is a hypothetical
scaling of the CTH model to actual test sled dimensions. The scaling is checked against invariant
products developed by applying the Buckingham Pi Theorem to the problem.
The dimensional analysis does not account for high-strain rate dependencies of the material
models. As such, numerical studies investigated strain rate sensitivity of the invariant products.
Strain rate differences between geometrically scaled models can be modified in accordance with
scaling laws by adjusting material constants in the constitutive equation for one of the models.
However, this technique is limited to constitutive equations which allow modification of constants
that affect strain rate. It was also shown that relative size of the mesh cells with respect to the
model and the number of cells defining the area of interest is an important factor for scaling
numerical models in CTH.
The invariant products developed in this investigation can be used as a set of qualitative
guidelines to aid in applying CTH model results to real test sleds. They also provide an indication
 ARTICLE IN PRESS

A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946 945

of what parameters of the CTH simulation model need changed in order to better represent the
particular dimensionality of a given test sled.

Acknowledgements

The authors would like to acknowledge Dr. Dean Mook from the Air Force Office of Scientific
Research for his generous financial support and Dr. Michael Hooser from the Holloman High
Speed Test Track for his invaluable technical expertise.

References

[1] Krupovage D, Rasmussen H. Hypersonic rocket sled development. Technical report JON:99930000, High Speed
Test Track Facility, 6585th Test Group, Holloman AFB, NM, September 1982.
[2] Krupovage D. Rail gouging on the holloman high speed test track. In: 35th meeting of the aeroballistic range
association, 6585th Test Group, Test Track Division, Holloman AFB, NM, 1984.
[3] Gerstle F. The sandia rocket sled rail and slipper study. Technical report Sandia National Laboratories,
Albuquerque, NM, 1968.
[4] Gerstle F. Deformation of steel during high velocity unlubricated sliding contact. Ph.D. thesis, Duke University,
1972.
[5] Gerstle F, Follansbee P, Pearsall G, Shepard M. Thermoplastic shear and fracture of steel during high-velocity
sliding. Wear 1973;24:97–106.
[6] Barber J, Bauer D. Contact phenomena at hypervelocities. Wear 1982;78:163–9.
[7] Tarcza K. The gouging phenomenon at low relative sliding velocities. Master’s thesis, The University of Texas at
Austin, Austin, TX, December 1995.
[8] Graff K, Dettloff B. The gouging phenomenon between metal surfaces at very high speeds. Wear 1969;14:
87–97.
[9] Graff K, Dettloff B, Bolbulski H. Study of high velocity rail damage. Technical report AA27 F29600-67-C-0043,
Air Force Special Weapons Center, Kirtland AFB, NM, August 1970.
[10] Hertel E, Bell R, Elrick M, Farnsworth A, Kerley G, McGlaun J, Petney S, Silling S, Taylor P, Yarrington L. Cth:
a software family for multidimensional shock physics analysis. Proceedings of the 19th international symposium
on shock waves I.
[11] Hooser M. Simulation of a 10,000 foot per second ground vehicle. In: 21st AIAA advanced measurement
technology and ground testing conference, no. AIAA 2000-2290, Denver, CO, 2000.
[12] Hooser M. Dynamic design and analysis system simulations. Unpublished data from Holloman AFB, NM, 2001.
[13] Laird D. The investigation of hypervelocity gouging afit/ds/eny 02-01. Ph.D. thesis, Air Force Institute of
Technology, Wright-Patterson AFB, OH, March 2002.
[14] Barker L, Trucano T, Munford L. Surface gouging by hypervelocity sliding contact. Technical report SAND87-
1328, Sandia National Laboratories, Albuquerque, NM, September 1987.
[15] Barker L, Trucano T, Susoeff A. Railgun rail gouging by hypervelocity sliding contact. IEEE Trans Magn
1988;25(1):83–7.
[16] Schmitz C, Palazotto A, Hooser M. Numerical investigation of the gouging phenomena within a hypersonic rail-
sled assembly. In: AIAA/ASME/AHS/ASC structures, structural dynamics, and materials conference and exhibit,
Seattle, WA, no. AIAA 2001-1191, 2001.
[17] Tachau R. An investigation of gouge initiation in high-velocity sliding contact. Technical report SAND91-1732,
Sandia National Laboratories, Albuquerque, NM, November 1991.
[18] Tachau R, Yew C, Trucano T. Gouge initiation in high-velocity rocket sled testing. Technical report SAND94-
1333C, Sandia National Laboratories, Albuquerque, NM, 1994.
 ARTICLE IN PRESS

946 A.G. Szmerekovsky et al. / International Journal of Impact Engineering 32 (2006) 928–946

[19] Laird D, Palazotto A. Gouge development during hypervelocity sliding impact. Int J Impact Eng 2004;30:205–23.
[20] Laird D, Palazotto A. Temperature effects on the gouging and mixing of solid metals during hypervelocity sliding
impact. In: 43rd AIAA structures, structural dynamics, and materials conference, Denver, CO, 2002.
[21] Laird D, Palazotto A. Effects of temperature on the process of hypervelocity gouging. AIAA J
2003;41(11):2251–60.
[22] Baker W, Palazotto A, Laird D. Thermal diffusion and associated stress field due to high speed source. ASCE J
Aerospace Eng 2002; 118–24.
[23] Jones N. Structural impact. New York, USA: Cambridge University Press; 1989. p. 489–519.
[24] Baker W, Westine P, Dodge F. Similarity methods in engineering dynamics, 2nd revised ed. vol. 12 of
Fundamental studies in engineering. New York, USA: Elsevier Science Publishing Company; 1991.
[25] Mullin S, Anderson C, Wilbeck J. Dissimilar material velocity scaling for hypervelocity impact. Int J Impact Eng
2003;29:469–85.
[26] Dancygier A. Scaling of non-proportional non-deforming projectiles impacting reinforced concrete barriers. Int J
Impact Eng 2000;24:33–55.
[27] Holsapple K. The scaling of impact phenomena. Int J Impact Eng 1987;5:343–55.
[28] Arione S, Bjorkman M. Scaling flow fields from the impact of thin plates. Int J Impact Eng 1987;5:61–7.
[29] McGlaun J, Thompson S, Elrick M. Cth: a three-dimensional shock wave physics code. Int J Impact Eng
1990;10:351–60.