You are on page 1of 17

Geotech Geol Eng

https://doi.org/10.1007/s10706-018-0659-8 (0123456789().,-volV)
(0123456789().,-volV)

ORIGINAL PAPER

Dynamic Properties of Geologic Specimens Subjected


to Split-Hopkinson Pressure Bar Compression Testing
at the University of Kentucky
Russell Lamont . Jhon Silva

Received: 23 February 2018 / Accepted: 1 August 2018


Ó Springer Nature Switzerland AG 2018

Abstract Advances in materials science have shown Sandstone, and Aluminum 6061-T6. Two of these are
that material behavior varies according to the rate of common aggregates found in the mining and con-
load application (strain-rate sensitivity). With regards struction industries, while the third is an aluminum
to compressive strength, materials have been observed variant often encountered in industrial applications.
to exhibit a strengthening, weakening, or negligible Dynamic compression testing of these materials at
response to increasing strain-rates (Zhang and Zhao in various strain rates was carried out, and the results are
Rock Mech Rock Eng 47(4):1411–1478, 2014). Prac- included. Static test results have been included for
tical experimentation to ascertain these responses has comparison, and the testing and data analysis proce-
been carried out for over a century, based on the dure are discussed in detail.
fundamental equipment design pioneered by John
Hopkinson in 1872 and modified by Kolsky in 1949. A Keywords Hopkinson bar  Compression testing 
contemporary Split-Hopkinson Pressure Bar (SHPB) Dynamic properties  Geologic samples  Strain
has been constructed at the University of Kentucky response
(UKY) to research the dynamic properties of various
geologic materials for mining and civil engineering
applications. Geologic samples are of an inconsistent
nature due to inherent discontinuities and large grain 1 Introduction
size. To ensure test specimens are of adequate size to
reflect this inconsistent nature, the SHPB at the UKY 1.1 Background
has been constructed with component bars of 2 in.
(5.08 cm) diameter. Prior publications have discussed The application of compressive loading is often
various considerations associated with the testing divided into two classifications: ‘‘static’’, and ‘‘dy-
procedure and data processing of this SHPB (disper- namic’’. As the names suggest, static loading repre-
sion correction, pulse shaping, etc.) (Silva and Lamont sents situations in which materials are subjected to
2017). This publication presents the results of mate- stationary loads which change negligibly over time. A
rials testing with this SHPB. Three materials were great amount of historical effort has gone into refining
selected: Bedford (Indiana) Limestone, Berea (Ohio) the static testing methodology, largely due to the
relative ease of data monitoring and processing that
R. Lamont (&)  J. Silva
accompanies a slow moving process. However, in
University of Kentucky, Lexington, KY, USA practical circumstances, real-world loads often change
e-mail: Russell.Lamont@uky.edu

123
Geotech Geol Eng

significantly over time. These are known as dynamic grade steel bars of 2 in. (5.08 cm) diameter. The
loads. The static properties of most materials are still incident and transmitted bars are 96 in. (2.45 m) long,
useful in indicating probable responses, even under while the striker bar (bullet) used for this testing is
dynamic loading, and thus are included in many 24 in. (0.61 m) in length with a mass of 20.28 lbf
engineering models. Attempts to more accurately (9.2 kg). The Poisson’s Ratio of this steel is 0.29, and
model and predict dynamic loading scenarios have led it possesses a density of 490.06 pcf (7.85 g/cc). A
to efforts to define dynamic properties, despite the compressed gas system capable of sustaining charge
degree of difficulty involved. pressures up to 700 psi (4.83 MPa) allows for control
The design and procedure of dynamic compression of impact velocity. Strain gauges are installed at the
testing is a complex task, as no single established locations shown in Fig. 1 above, with multiple gauges
standard for this testing has yet attained general at each location to allow for signal combination, thus
acceptance. Instead, the science describing the phys- reducing any signal variation possible from gauge
ical processes which occur during dynamic testing has installations. An MREL DataTrapII data recorder is
been explored and explained, which may then be used used to capture the signals, with sampling rates of up
by researchers to establish independent but often to 10 MSa/s. More detailed discussion of the SHPB
similar test protocols. The construction and operation characteristics and experimental setup has been cov-
principles of the SHPB at the UKY followed basic ered by Silva and Lamont (2017) (Fig. 2).
guidelines which are summarized by the following The pulse generated by the impact of the striker on
characteristics: the incident bar will propagate at the wave speed of the
bar material cb until reaching the interface of the
1. Two equal-length, thin bars (known as the inci-
incident bar and test specimen. At this point, a portion
dent and transmitted bars) constructed of a high-
of the strain pulse will be reflected back and a portion
strength material such as steel, titanium, or nickel
will be transmitted through the specimen. This concept
alloy.
is depicted in Fig. 3.
2. A striker bar constructed of the same material,
Despite equipment design differences and varying
typically designed to be the same diameter but of
applications, the basic science of the SHPB apparatus
much shorter length than the other two bars. Using
at the UKY remains similar to that explored by other
a barrel, this striker bar is directed to strike in
researchers. Thus, derivation of the dynamic formulas
alignment with the incident bar.
which dictate determination of sample compressive
3. A compressed gas system with quick-release
strength is essential to understanding the experimental
valve used to impart momentum to the striker
process. Frew et al. (2002) provide an excellent
bar, and a laser interruption type device used to
discussion of this derivation, an adaptation of which is
capture striker velocity.
given in this section. Note that the component steel
4. Minimum of two strain gauges bonded to the
bars of the SHPB possess properties of density q,
surface of the incident and transmitted bars and
Young’s Modulus Eb , bar wave speed cb , and cross-
connected to a data acquisition system. These
sectional area Ab . When referring to the directions
strain gauges are usually installed equidistant
shown in Fig. 3, the derived equations follow sign
from the specimen-bar interfaces to simplify the
conventions of positive stress in compression, positive
data processing.
strain in contraction, and positive particle velocity to
As shown in Fig. 1, the SHPB constructed at the the left. Subscripts 1 and 2 represent the incident-
UKY is composed of perfectly machined SAE 4340 specimen and specimen-transmitted interfaces,

Fig. 1 Configuration and


dimensions of the SHPB at
the UKY (U.S. customary
units)

123
Geotech Geol Eng

Fig. 2 Photographs of the SHPB at the UKY and the placement of the material specimen within the apparatus

Fig. 3 Diagram showing a


general schematic of the
SHPB device along with
direction and nomenclature
of the induced strain waves
adapted from Gray III
(2000)

respectively. Subscripts i, r, and t, denote the incident, unlikely that the two values will ever reach precisely
reflected, and transmitted strain waves as shown in the same value for any appreciable length of time, but
Fig. 3. If the material specimen being tested has an in general, a level of variation may be defined as
initial cross-sectional area As and length Ls , then the acceptable to the experimental setup. For the exper-
stress at interfaces 1 and 2 may be calculated as iments described in this paper, that level is defined as
Eb Ab 5%. Once this level has been achieved, the stress (rs ),

r1 ¼ ð ei þ er Þ ð1Þ strain rate dedts , and strain ðes Þ of the specimen may be
As
calculated using either the 1-wave or 3-wave
Eb Ab approaches (Lu and Li 2010). Equations 3 to 5
r2 ¼ ð et Þ ð2Þ
As representing the 3-wave approach are commonly used
when the attainment of stress equilibrium is unknown
As the incident strain pulse reaches the specimen, a
or difficult to assess.
portion of the pulse will undergo reflections within the
specimen, due to the impedance mismatch which Ab
rs ðtÞ ¼ E b ½ ei ð t Þ þ er ð t Þ þ e t ð t Þ  ð3Þ
exists at interfaces 1 and 2. As these reflections take 2As
place, the specimen gradually reaches dynamic stress
des ðtÞ cb
equilibrium. Thus, as time elapses in the experiment, ¼ E b ½ ei ð t Þ  er ð t Þ  et ð t Þ  ð4Þ
the difference between the values of r1 and r2 should dt Ls
decrease and ultimately reach equivalency. In reality,
due to various loading and wave phenomena it is

123
Geotech Geol Eng

Table 1 Listing of various static properties for the materials tested in this research. (Hill 2017; Churcher et al. 1991; ASM
International 1990; Holt 1996)
Limestone Sandstone Aluminum 6061-T6

Compressive strength 4000? psi (27.58? MPa) 6150–8670 psi (42.40–59.78 MPa) 35,000? psi
(241.32? MPa)
Specific gravity 2.10–2.75 2.06 2.70
Modulus of elasticity 3300–5400 kpsi 2010–2270 kpsi 10,000 kpsi (68,948 Mpa)
(22,753–37,232 MPa) (13,858–15,651 MPa)
Shear strength 900–1800 psi (5.52–12.41 MPa) 1160–5802 psi (8–40 MPa) 30,000 psi (207.84 MPa)
Ultimate tensile 300–715 psi (2.07–4.93 MPa) 580–3626 psi (4–25 Mpa) 45,000 psi (310.26 MPa)
strength
Poisson’s ratio 0.18–0.33 0.21–0.38 0.33

cb t this material. The descriptions of the materials and


es ðt Þ ¼ r ½ei ðtÞ  er ðtÞ  et ðtÞ ð5Þ their static properties can be found in several publi-
Ls 0
cations and are listed in Table 1.
Once stress equilibrium is reached, the relationship Bedford (Indiana) Limestone Mississippian-age
given by ei ðtÞ þ er ðtÞ ¼ et ðtÞ is true, and the equations free-stone (homogenous texture and grade). This stone
will reduce to the 1-wave solution given in Eqs. 6 to 8. is considered chemically pure, with 97% calcium
This emphasizes the importance of specimen stress carbonate and 1.2% calcium–magnesium carbonate.
equilibrium (Wu and Gorham 1997). In this research Indiana Limestone has been used extensively in
both the 1-wave and 3-wave equations were used in building construction, including the Empire State
order to allow comparison between the two solutions. Building.
Eb A b Aluminum 6061-T6 Commonly used industrial
rs ðtÞ ¼ et ðt Þ ð6Þ
As aluminum alloy. Typical applications include aircraft
fittings, couplings, marine fittings and hardware,
des ðtÞ 2cb pistons, bike frames, and many others. Generally
¼ er ðt Þ ð7Þ
dt Ls recognized to exhibit excellent joining characteristics,
good acceptance of applied coatings, relatively high
2cb t
es ðt Þ ¼ r er ðsÞds ð8Þ strength, good workability, and high resistance to
Ls 0 corrosion.
Berea (Ohio) Sandstone Mississippian-age rock
composed primarily of quartz (87–97%) with trace
2 Testing Procedure amounts of feldspar. Commonly referred to as ‘‘Berea
Grit’’ by the oil and gas industry. Ohio Sandstone is a
2.1 Sample Materials and Data moderate to well sorted, medium-grained, quartz-rich
sandstone. Matrix content (silt and clay) is minimal,
Three common materials were selected for dynamic making up 2% or less of the rock.
compression testing using the SHPB at the UKY.
For the first two materials, past published research 2.2 Specimen Preparation and Geometry
was found which allowed for comparison of results. Selection
These results will be presented as verification that the
SHPB construction and data analysis procedure reflect The sample preparation and geometry selection for
past research findings. Literature containing dynamic static compression testing was performed according to
properties for the third material (Berea Sandstone) ASTM standards D4543 and D7012. A length to
could not be found, and thus the results are included as diameter ratio of 2:1 was selected for the specimens
a viable characterization for the strain-rate response of subjected to static tests and the results obtained from
these 2 in. (5.08 cm) diameter samples are presented.

123
Geotech Geol Eng

The values obtained here provide a baseline reference that a length should be selected such that it yields a
to which other experimental results may be compared, length to diameter ratio within the range of 0.5 to 2.0
as the dynamic testing requires a different sample (Gray III 2000; Gama et al. 2004). According to the
geometry. These size and shape effects have been authors, one of the most fundamental requirements of
explored in depth elsewhere (Broch 1983; Forster the specimen length is that the specimen should be
1983; Petrov and Selyutina 2015). short enough to ensure the sample reaches stress
Selection of the dynamic specimen geometry equilibrium in a timely manner, and yet still long
involves several considerations not specified by enough to limit axial inertial effects.
ASTM standards, each of which has the potential to Figure 4 shows actual experimental results
affect the results obtained during testing. First, an obtained using the SHPB at the UKY for the face
appropriate sample diameter must be selected. As the stress values of r1 and r2 over time. The initial
bars of this SHPB are constructed of 2 in. (5.08 cm) ‘‘ringing-up’’ period at the beginning of the experi-
diameter steel, it is important that the diameter of the ment is easily identified, and consists of the minimum
specimens be less than this dimension. Some 3–4 pulse reflections required to reach equilibrium in
researchers cite 80–90% of the bar diameter as an the sample (Davies and Hunter 1963; Ravichandran
appropriate specimen diameter (Gama et al. 2004). and Subhash 1994). Once the specimen reaches
Following this recommendation, a specimen diameter equilibrium, the two values nearly approximate one
of approximately 1.75 in. (4.45 cm) was chosen for another until the failure process begins to occur. A
dynamic testing. Next, the length of the samples had to variance between the values of less than 5% may be
be selected. When dynamic compression testing using considered acceptable (Subhash and Ravichandran
the SHPB, no definite standard exists for the selection 2000). Wu and Gorham (1997), and Gong et al. (1990)
of specimen length. Instead, some researchers state have published similar results.
The data obtained during the ‘‘ring-up’’ period of
wave reflections should not be regarded as valid, and if
Geologic Specimen Face Stress Values failure occurs during this stage the test results should
11000

σ1 be discarded. Although in this particular figure r1


10000 σ2 shows as initially negative, this is not always the case,
σ1, Average of 10 points
σ2, Average of 10 points as shown in Fig. 11. The data obtained during this
9000
stage is erroneous due in large part to impedance
8000 mismatch, inertial effects, and the physical impossi-
bility of ensuring a seamless connection between the
7000
bars and specimen. The time required to achieve
6000 equilibrium (sm Þ may be easily calculated using the
length of the specimen Ls , the elastic wave speed in the
Stress (psi)

5000
specimen cs , and the minimum number of wave
4000
reflections required. According to literature, the rec-
ommended wave reflections are between 3 and 4
3000
(Davies and Hunter 1963), and possibly as much as 6–
2000
8 (Ravichandran and Subhash 1994). The travel time
ðto Þ from one bar-specimen interface to the other is
1000
given by Chen et al. (1999):
0 Ls
to ¼ ð9Þ
cs
-1000

Once this travel time has been determined, the


-2000
439.24 439.26 439.28 439.3 439.32 439.34 439.36 439.38 439.4 439.42 439.44
number of stress wave reflections ðnÞ necessary to
Time (milliseconds)
ensure dynamic equilibrium can be calculated using
Fig. 4 Values of r1 and r2 over time in an actual SHPB the expression given by:
experiment performed on a limestone sample at the UKY

123
Geotech Geol Eng

 
sm 2.3 Static Testing Procedure
n¼ ð10Þ
2t0
Static testing of the various materials was carried out
where ðsm Þ is the minimum required loading time. using a SATEC C600B Compressive Testing System
This yields that the required rise time of an incident with Instron controller. Testing was performed on
pulse must be greater than sm . For the dynamic SHPB samples with the typical 2:1 length to diameter ratio,
experiments presented in this paper, a length to as well as samples having a 1:1 ratio, to isolate the
diameter ratio of 1:1 was selected. This meant the effect of the length to diameter ratio variance. This
limestone and sandstone specimens were ground to a effect is shown and discussed in the results portion of
length of approximately 1.75 in. (4.45 cm). Using the this paper. The photographs in Fig. 6 depict the static
knowledge that the Berea Sandstone is the most testing of 2:1 ratio rock samples, 1:1 ratio rock
porous of the materials tested and hence has the lowest samples, and the before and after condition of the 1:1
wave velocity of 7808 ft/s (2380 m/s) (Shankland aluminum samples (notice the barreling-type failure as
et al. 1993), it is possible to find the maximum required opposed to the brittle failures of the rocks).
loading time for the experiments carried out with this
specimen length. To guarantee dynamic equilibrium, 2.4 Dynamic Testing Procedure
it was assumed that 8 reflections are required. Thus,
the wave must travel a distance of 1.75 in. 9 8 = 14 The dynamic testing procedure involving the UKY-
in. (35.56 cm). The time required to do so is 149 SHPB was carried out on the sandstone, limestone,
microseconds. Comparing this time to the pulse- and aluminum samples with 1:1 length to diameter
shaped waveform shown in Fig. 7, it appears that ratios. A Ranger high-speed camera was used to
theoretically even the ‘‘worst case’’ requirements capture the events visually at frame rates of up to
(defined as requiring the greatest span of time to 16,000 fps, as the event duration for each test may be
achieve equilibrium) should still be met by the measured in microseconds. A planned test matrix was
constant-rate loading portion of the shaped pulse. designed in which compression tests for each of the
In contrast to the rock specimens, the aluminum three materials would be carried out at system charge
samples were obtained from a round bar already pre- pressures of 20, 40, 60, 120, 240, and 400 psi (138,
machined to the required diameter. The bar was then 275, 413, 827, 1654, 2758 kPa). It was anticipated that
cut at intervals to produce a 1:1 length to diameter this change in charge pressure would provide some
ratio. Figure 5 below provides a visual representation variation in the experimental strain rate, which could
of the rock and aluminum samples, along with a side- be used to determine the material’s strain rate
by-side comparison of the size difference between the response. Strain gauges manufactured by Micro-
samples prepared for traditional static testing and Measurements with resistances of 120 ohms in the
those prepared for dynamic SHPB testing. CEA series were selected with grid dimension
constraints allowing for data recording at rates

Fig. 5 Photographs showing the rock and aluminum test specimens, at the geometries selected for both static and dynamic testing

123
Geotech Geol Eng

Fig. 6 Photographs showing the static testing carried out on both geologic and aluminum samples of 1:1 and 2:1 geometric ratios

exceeding the maximum recording rate of the capture 1750

device (5 MS/s). To limit the effects of radial inertia, a Shaped


silicone-based lubricant was applied to the specimen-
bar interfaces. The data was recorded at sample rates
1250
of 5 Ms/s, and a 0.75 in. (1.91 cm) diameter by 16 GA
thick copper pulse shaper was used for each test. A
pulse shaper is a small disc, often composed of a soft
Microstrains

metal, which through deformation lengthens the pulse 750


load time. It is placed on the end of the incident bar
impacted by the striker, and upon impact will deform,
slowing the transfer of energy. This is often necessary
250
to ensure the constant-rate loading phase of the pulse is
adequate to achieve stress equilibrium. Prior testing of
a wide range of disc materials and geometries yielded
the conclusion that this particular pulse shaper gener- -250
280 380 480 580 680 780
ates the most desirable pulse results. The pulse
Microseconds
generated by using this particular shaper with no test
specimen between the incident and transmitted bars is Fig. 7 Graph showing the triangular pulse obtained by using a
shown in Fig. 7 (the thick dashed line). The unshaped 0.75 in. (1.91 cm) by 16 GA copper pulse shaper, compared to
the unshaped waveform at the UKY–SHPB
pulse (thin solid line) is also shown for comparison.
Note the much longer rise time with a relatively
Still frames captured by the hi-speed camera are
constant slope, the length of which exceeds the
shown in the photographs of Fig. 8, both pre- and post-
‘‘worst-case’’ requirement of 149 microseconds. Thus,
failure. Note the ‘‘shattering’’ type failure experienced
the shaped pulses used for this testing are adequate to
by the brittle rock specimens, as opposed to the
allow achievement of stress equilibrium of the spec-
relatively indiscernible shape variation of the alu-
imens in even the most extreme situations. Other
minum specimens (some barreling effect was observ-
benefits of using the pulse shaper include the reduction
able under close inspection).
of oscillations in the loading phase and dampening of
dispersion effects.

123
Geotech Geol Eng

Fig. 8 Photographs showing the pulse wave effects on the rock and aluminum samples

3 Results may be attributed to the fact that rock types exhibiting


extremely homogenous and uniform structure were
3.1 Static Testing Results and Discussion of Data selected for testing, and thus any effect of geologic
Processing discontinuities may be insignificant. Uniformity of the
aluminum is believed to be the main explanatory
Static testing was performed for each of the three test factor behind the negligible change in failure strength
materials at both the 2:1 and 1:1 geometry ratios. Up to as well.
8 individual tests were run for each of 6 possible Valuable information obtained from this testing
experimental scenarios, and the average results are included the average failure strengths of the 1:1
depicted in the three graphs below (Fig. 9). Note that samples, which provide the starting point (origin) for
in all scenarios, the stress–strain curve shifts to the strain response curves which were later completed
right when reducing the ratio from 2:1 to 1:1. This may using the data obtained during dynamic testing. These
be translated as a reduction in the Young’s Modulus of average failure points are represented in Table 2.
the material (represented by the slope of the linear
portions of the curves). However, only negligible
differences were observed in the yield strength. This

a b c
Static Uniaxial Compressive Stress-Strain Static Uniaxial Compressive Stress- Static Uniaxial Compressive Stress-Strain
Average Curve Bedford Limestone Strain Average Curve Aluminum 6061-T6 Average Curve Ohio Sandstone
7000 48.26 60000 413.69 8500 58.61
L:D = 2:1 L:D = 2:1 L:D = 2:1
6500 L:D = 1:1 44.82 8000 L:D = 1:1 55.16
55000 L:D = 1:1 379.21
7500 51.71
6000 41.37
50000 344.74 7000 48.26
5500 37.92 6500 44.82
45000 310.26
5000 34.47 6000 41.37
40000 275.79
Stress (MPa)
Stress (MPa)

Stress (MPa)

5500 37.92
Stress (psi)

4500 31.03
Stress (psi)

Stress (psi)

35000 241.32 5000 34.47


4000 27.58
4500 31.03
3500 24.13 30000 206.84
4000 27.58
3000 20.68 25000 172.37 3500 24.13
2500 17.24 3000 20.68
20000 137.90
2000 13.79 2500 17.24
15000 103.42 2000 13.79
1500 10.34
10000 68.95 1500 10.34
1000 6.89
1000 6.89
500 3.45 5000 34.47
500 3.45
0 0.00 0 0 .0 0 0 0.00
0 1000 2000 3000 4000 5000 6000 0 15000 30000 45000 60000 0 4000 8000 12000

Strain (microstrains) Strain (microstrains) Strain (microstrains)

Fig. 9 Graphs showing the quasi-static stress–strain average curves for the three test materials, at geometric ratios of 2:1 and 1:1

123
Geotech Geol Eng

Table 2 Average quasi-static failure strengths of the test Specimen Face Stresses
materials, at 2:1 and 1:1 geometric ratios Ohio Sandstone - 20psi
16000 110.32
Static failure strengths σ1
σ2
Material L:D ratio r (psi) 14000 96.53

Indiana limestone 2:1 6562


12000 82.74
Indiana limestone 1:1 6246
Ohio sandstone 2:1 7774
Ohio sandstone 1:1 8038 10000 68.95

Aluminum 6061-T6 2:1 45,821

Stress (MPa)
Stress (psi)
Aluminum 6061-T6 1:1 46,105 8000 55.16

6000 41.37

3.2 Discussion of Data Processing


4000 27.58

The dynamic testing procedure is much more complex


than the static, largely due to the lack of standardized 2000 13.79

equipment and data processing tools. To demonstrate


the data processing procedure followed in this 0 0.00
research, Figs. 10, 11, 12, 13 and 14 were derived
-2000 -13.79
450.25 450.4 450.55 450.7
Dynamic Strain-Time Curves Time (milliseconds)
Ohio Sandstone - 20psi
1000
Fig. 11 Calculated stress levels on the faces of an Ohio
Sandstone test specimen, subjected to a 20 psi (138 kPa) charge
800
impact

and included from a test of Ohio Sandstone at a charge


600 pressure of 20 psi (138 kPa). Although not shown in
Fig. 10, the first step in the data reduction process is
400
the averaging of signals from strain gauges installed
Strain (microstrains)

equidistant on the bars (i.e. 2 gauges were installed at


the midpoint of both the incident and transmitted bars,
200 and thus the signals shown in Fig. 10 are the average
of these two signals at each midpoint).
0
Once these average signals have been obtained, it is
necessary to time-shift the incident wave in such a way
that it is superimposed over the transmitted and
-200 reflected waveforms, with a relevant common time
Incident/Reflected
Transmitted origin. To perform this action, an initial starting point
Incident (shifted) of the incident wave must be defined. For the analyses
-400
performed in this paper, the starting points of the
incident waves were defined by finding the slope of
-600
449.6 449.8 450 450.2 450.4 450.6 450.8 data in ‘‘windows’’, containing 200 individual data
Time (milliseconds) points. These slopes were then analyzed to find the
coefficient of variation for these windows, using
Fig. 10 Graph showing the incident (positive), transmitted and Eq. 11:
reflected (negative) waves, the time-shifted incident wave is
included

123
Geotech Geol Eng

Fig. 12 a Graph showing a Specimen Stress-Strain Curve b Specimen Stress-Strain Curve


the stress–strain curve for a Ohio Sandstone - 20psi Ohio Sandstone - 20psi
1 600 0 110.32 14300 98.60
test of Ohio Sandstone Average of 10 points Average of 10 points
14200 97.91
subjected to a 20 psi
14000 96.53 14100 97.22
(138 kPa) charge, and
b enlargement of a portion 14000 96.53
12000 82.74
of the same curve important 13900 95.84
to failure determination 13800 95.15
10000 68.95
13700 94.46

Stress (MPa)
Stress (MPa)

Stress (psi)
Stress (psi)
8000 55.16 13600 93.77
13500 93.08
6000 41.37 13400 92.39
13300 91.70
4000 27.58
13200 91.01
13100 90.32
2000 13.79
13000 89.63

0 0.00 12900 88.94


12800 88.25
-2000 -13.79 12700 87.56
0 4000 8000 12000 16000 20000 13800 14700 15600 16500
Strain (microstrains) Strain (microstrains)

Slope of Specimen Stress-Strain Curve Slope Rate of Change of Stress-Strain Curve


Ohio Sandstone - 20psi Ohio Sandstone - 20psi
2 0.005
1.8 δ/δx(δ/δx(Average of 10 points))

1.6 0.004
Slope Rate of Change of Stress-Strain Curve

1.4
1.2 0.003
1
Slope of Stress-Strain Curve

0.8 0.002
0.6
0.4 0.001
0.2
0 0
-0.2
-0.4 -0.001
-0.6
-0.8 -0.002
-1
δ/δx (Average of 10 points)
-1.2 -0.003
-1.4
-1.6 -0.004
-1.8
-2 -0.005
13800 14400 15000 15600 16200 16800 13800 14400 15000 15600 16200 16800
Strain (microstrains) Strain (microstrains)

Fig. 13 a Graph showing the first derivative (slope) of the stress–strain curve from the Ohio Sandstone 20 psi (138 kPa) test, and
b graph showing second derivative (rate of change of the slope) of the same stress–strain curve

r be attributed to random ‘‘static’’, the limit of


CoV ¼  100% ð11Þ
l CoV = 50 provides a standardized means of designat-
ing the start of the waveform. This method and limit
A CoV limit of 50 was deemed to be most
was applied to each test run.
representative of the graphically verifiable starting
Some authors have recommended using the point at
point. Thus, once the recorded signal enters an area
which the incident wave crosses the x-axis as the start
where the slope is consistently trending positively as
of the reflected wave (Kaiser 1998). However, in the
opposed to the positive–negative variation which may
course of testing at the UKY it was found that trials

123
Geotech Geol Eng

1 & 3-Wave Solutions and Strain Rate incident wave signal crosses the x-axis (this effect will
Ohio Sandstone - 20psi
17500 210 be discussed in more detail later in this section). Due to
1-Wave this stress bleed-off condition, the reflected wave
3-Wave
Strain Rate starting point was defined using the speed of pulse
15000 180
travel in the steel bars (found in previously published
experiments to be 5148.6 m/s or 16,891.7 in./s), to
12500 150 avoid dependence of the reflected wave starting point
on the incident wave ending point. Coupling this
10000 120 velocity with the knowledge that the strain gauges are

Strain Rate (s-1)


placed at the midpoints of the bars 48 in. (1.22 m), the
Stress (psi)

7500 90
elapsed time required for the signal to travel the 96 in.
(2.44 m) from the strain gauge to the specimen and
back was found. Thus the incident wave may be
5000 60
superimposed over the reflected and transmitted waves
by adding this time (found to be 473.60 ls). In this
2500 30 way, the time-shifted incident wave in Fig. 10 was
generated and placed on the x-axis.
0 0 In Fig. 11, a representation of the stress on each of
the specimen faces over the elapsed time of the
experimental event is shown. The values graphed here
-2500 -30
0 4000 8000 12000 16000 20000 are obtained by application of Eqs. 1 and 2 to the
Strain (microstrains) recorded signals. Note that for the portion of the
Fig. 14 Graph showing the stress values for the 1 and 3-wave
experiment which is pertinent (before specimen fail-
solutions, as well as the strain rate of the experiment ure), the stresses on the two faces are roughly
equivalent. Thus, we may conclude that the specimen
using high impact energy levels (high strain rates), was in dynamic stress equilibrium when failure
presented a ‘‘stress bleed-off’’, causing the trailing occurred. During the data analysis process, it was
edge of the incident signal to be affected. This found that for high energy impacts (those with charge
condition significantly altered the point at which the pressures of 120 psi (827 kPa) and over), it was

Fig. 15 a Graph showing a Dynamic Strain-Time Curves b Specimen Face Stresses


the incident (positive Indiana Limestone - 400psi Indiana Limestone - 400psi
5000 9.5E+10 6.55E+8
phase), reflected (negative
9E+10 6.21E+8
phase), transmitted, and 8.5E+10 5.86E+8
4000
time-shifted incident waves 8E+10 5.52E+8
7.5E+10 5.17E+8
for a test conducted on 7E+10 4.83E+8
3000
Indiana Limestone at 6.5E+10 4.48E+8
400 psi (2758 kPa), and 6E+10 4.14E+8
Strain (microstrains)

2000 5.5E+10 3.79E+8


b graph showing that the 5E+10 3.45E+8
Stress (MPa)

specimen face stresses of


Stress (psi)

4.5E+10 3.10E+8
1000 4E+10 2.76E+8
this test conducted on
3.5E+10 2.41E+8
Indiana Limestone at 3E+10 2.07E+8
0
400 psi (2758 kPa) never 2.5E+10 1.72E+8
2E+10 1.38E+8
approach one another 1.5E+10 1.03E+8
-1000
(dynamic equilibrium never 1E+10 6.89E+7
achieved) 5E+9 3.45E+7
-2000 Incident/Reflected 0 0.00
Transmitted -5E+9 -3.45E+7
Incident (shifted) -1E+10 σ1 -6.89E+7
-3000 -1.5E+10 σ2 -1.03E+8
-2E+10 -1.38E+8
-2.5E+10 -1.72E+8
-4000 -3E+10 -2.07E+8
127 127.2 127.4 127.6 127.8 128 128.2 127.55 127.7 127.85 128
Time (milliseconds) Time (milliseconds)

123
Geotech Geol Eng

difficult if not impossible to achieve this stress meaningful, up to the 60 psi (414 kPa) charge pres-
equilibrium. This is likely due to the specimen length sure. For the trials using pressures of 120, 240, and
being too great to allow achievement of stress 400 psi (827, 1654, 2758 kPa), a stress ‘‘bleed-off’’
equilibrium before failure occurs. Future testing with tail appeared on the trailing edge of the incident wave.
shorter specimens will explore this hypothesis. This significantly impacts the values of the reflected
Using the superimposed incident, transmitted, and wave, and hence the results were rendered invalid and
reflected waves, the 1-wave and/or 3-wave solution discarded. This stress bleed-off is visible in Fig. 15a.
equations (Eqs. 3 to 8) may be used to determine the Note how the trailing edge of the incident signal bleeds
average stress, strain, and strain rate at any point in over the transmitted wave, and appears to affect the
time. Once this data is obtained, the familiar stress– leading edge of the reflected wave. This figure comes
strain curve may be generated, as shown in Fig. 12. from an experiment run on Indiana Limestone at a
Note that there is some level of ‘‘static’’, or abrupt 400 psi (2758 kPa) charge pressure. Additionally,
variation between each data point. To reduce this Fig. 15b shows how for the same test, the specimen
unwanted variation while still maintaining data accu- face stresses never approach equivalency. Similar
racy, a 10 point averaging function was carried out, as effects have been reported by another researcher when
shown in Fig. 12a. This is a necessary first step in high-velocity tests were carried out (Wu and Gorham
finding the precise failure point in the data. Visual
inspection of the graph allows easy determination of Bedford (Indiana) Limestone Stress-Strain Curves
which part of the signal is pertinent to failure analysis, 14000 96.53
and an enlarged portion of the averaged waveform in
Fig. 12a is shown in Fig. 12b. 13000 89.63

One technique for finding the precise failure point is Strain Rate 158s -1
12000 82.74
finding the point of inflection. This means finding the Strain Rate 150s-1
point at which the slope of the stress–strain curve 11000 Strain Rate 155s-1 75.84
changes most rapidly (in the negative direction). Strain Rate 149s -1
10000 -1
Strain Rate 105s68.95
Figure 13a is the graphical representation of the first
derivation of the enlarged portion of the stress–strain Strain Rate 95s-1
9000 62.05
curve in Fig. 12b. As expected, this line trends
downward over time, as the slope of the stress–strain 8000 55.16
curve in this area changes from positive to negative.
7000 48.26
Figure 13b is the graphical representation of the

Stress (MPa)
Stress (psi)

second derivation, which is the rate of change of the 6000 41.37


Static
slope of the stress–strain curve. Failure point deter-
mination corresponds with that part of the curve which 5000 34.47

changes most rapidly in the negative direction. The


4000 27.58
most negative point of the second derivative is visible
in this figure. The strain value at which this occurs may 3000 20.68
be transferred to the time domain, allowing for simple
2000 13.79
correlation to the stress value at which failure occurs.
Another verification of the specimen stress equi- 1000 6.89
librium is a comparison of the solutions given by the
1-wave and 3-wave methodologies. Figure 14 depicts 0 0.00
the stress value solutions of these two methods. Note
-1000 -6.89
the nearly equivalent values. Also shown on this graph
is the curve of the strain rate solution. The plateau -2000 -13.79
-2000 4000 10000 16000
portion of this curve is used to determine the strain rate
Microstrains
of the experiment.
Carrying out this analysis procedure for each of the Fig. 16 Graph of the stress–strain curves for each of the tests
experimental trials yielded results that appeared to be conducted on Bedford (Indiana) Limestone

123
Geotech Geol Eng

1997). Future testing will address this issue by finding


ways to control the stress bleed-off, such as using
shorter length striker bars, etc.

4 Dynamic Testing Results and Comparisons

4.1 Bedford (Indiana) Limestone

Results from dynamic testing of the Bedford Lime-


stone are shown in Figs. 16 and 17. The average result
from the static testing is included in each figure for
comparison. A substantial degree of scatter is evident
in the results, but in general a strong trend is
manifested of higher strain rate leading to higher
failure strength. The individual stress–strain curves of
each limestone test are shown in Fig. 16. Figure 17 is
a plot of the relationship between the failure strength Fig. 18 Graph showing the stress strain curves from testing
and strain rate of each test. Note that the overall carried out on Aluminum 6061 T-6
positive trend is highly significant for the range of
strain rates tested. and the results are shown in Fig. 17b. The results of
Dynamic testing of the Bedford (Indiana) Lime- this testing largely confirm the results obtained by the
stone was previously carried out by Frew et al. (2001) SHPB at UK. Interestingly, those authors also tested

Bedford (Indiana) Limestone


a 14000 96.53
R2 = 0.941
13600 93.77 b
13200 91.01

12800 88.25

12400 85.50

12000 82.74

11600 79.98

11200 77.22
Failure Strength (MPa)
Failure Strength (psi)

10800 74.46

10400 71.71

10000 68.95

9600 66.19

9200 63.43

8800 60.67

8400 57.92

8000 55.16

7600 52.40

7200 49.64

6800 46.88

6400 44.13

6000 41.37
1 2 3 4 5 7 10 20 30 50 100 200
Strain Rate (s^-1)

Fig. 17 a Plot of the Bedford (Indiana) Limestone failure strength versus the strain rate of each test., and b graph showing the results of
dynamic testing on Bedford (Indiana) Limestone performed by and figure taken from Frew et al. (2001)

123
Geotech Geol Eng

specimens exhibiting length to diameter ratios from 1 until reaching a strain rate of 103 s-1, this aluminum
to 2, and reported no significant effect. This serves as alloy does not exhibit strain rate sensitivity. Because
confirmation in the dynamic range of the static range the range of values tested at UK was at a maximum of
geometric ratio testing carried out at the UKY. approximately 150 s-1, this explains why no signif-
icant change in yield strength was observed. If future
4.2 Aluminum 6061 T6 testing were carried out a higher strain rates, confir-
mation of the higher range results obtained by Manes
Results from testing carried out on Aluminum 6061 et al. (2011) could be obtained.
T-6 are shown in Figs. 18 and 19a. Note that although
it is clear that the Young’s Modulus shifted to the left 4.3 Berea (Ohio) Sandstone
for each test, no discernible effect on failure strength
was observed. Also, it appears that the test conducted Figures 20 and 21a show the results of testing
with a 20 psi (138 kPa) impact charge (51 s-1 strain performed on the Berea (Ohio) Sandstone. Although
rate) may not have possessed enough energy to induce it was clear from testing that all dynamic testing
failure in the aluminum sample. results yielded failure strengths above the static testing
Dynamic testing of Aluminum 6061 T-6 was average, it is not apparent that within the range of
previously carried out by Manes et al. (2011) over a dynamic strain rates tested that increasing the strain
wide range of strain rates, from quasi-static up to rate had any effect on the strength. This is evidenced
104 s-1. The results of this testing are shown in by the random scatter of the stress–strain curves
Fig. 19b. Conclusions from this research were that

a Aluminum 6061 T-6


60000 413.69
2
R = 0.003
58000 399.90

56000 386.11 b
54000 372.32

52000 358.53

50000 344.74

48000 330.95

46000 317.16
Failure Strength (MPa)
Failure Strength (psi)

44000 303.37

42000 289.58

40000 275.79

38000 262.00

36000 248.21

34000 234.42

32000 220.63

30000 206.84

28000 193.05

26000 179.26

24000 165.47

22000 151.68

20000 137.90
1 2 3 4 5 7 10 20 30 50 100 200
Strain Rate (s^-1)

Fig. 19 a Graph showing the plot of strain rate versus failure strength for tests carried out on Aluminum 6061 T-6, and b graph showing
the results of dynamic testing on Aluminum 6061-T6 performed by and figure obtained from Manes et al. (2011)

123
Geotech Geol Eng

Berea (Ohio) Sandstone Stress-Strain Curves 5 Discussion and Conclusions


16000 110.32

Strain Rate 182s-1


15000 103.42 In this work testing of three materials was carried out
-1
using an SHPB of 2 in. (5.08 cm) diameter con-
14000 Strain Rate 185s96.53
structed at the University of Kentucky. The materials
13000 Strain Rate 85s-1 89.63 tested were Bedford (Indiana) Limestone, Berea
Strain Rate 94s-1 82.74
(Ohio) Sandstone, and Aluminum 6061-T6. The tests
12000
were carried out using samples with 1:1 length to
-1
75.84
11000 Strain Rate 210s diameter ratios, at strain rates ranging from quasi-
10000 68.95 static to 210 s-1.
Conclusions of this work are as follows.
9000 62.05

Static 1. The limestone samples exhibit a strong positive


8000 55.16
correlation between strain rate and failure

Stress (MPa)
Stress (psi)

7000 48.26 strength. This is confirmed by work previously


carried out by Frew et al. (2001).
6000 41.37
2. The aluminum samples exhibit no significant
5000 34.47 strain rate sensitivity for the range tested. This is
27.58
confirmed by work previously carried out by
4000
Manes et al. (2011). Future testing at higher strain
3000 20.68 rates may allow for confirmation of the positive
2000 13.79 correlation at high rates shown in that research.
3. The sandstone samples exhibit a weak positive
1000 6.89
correlation between strain rate and failure
0 0.00 strength. Without other published research for
this specific rock, this is difficult to confirm.
-1000 -6.89
However, work performed by Liu et al. (2011) on
-2000 -13.79 another sandstone did show a strong positive
-2000 6000 14000 22000 30000
Microstrains
correlation. Additional testing is recommended to
confirm the results from this publication.
Fig. 20 Graph showing the stress–strain curves for dynamic 4. Future testing should be performed with attempts
testing carried out on sandstone specimens to control the ‘‘stress bleed-off’’ phenomena
which caused the higher velocity (and higher
shown in Fig. 20, and by the weak positive correlation strain-rate) experimental data to be discarded.
plotted in Fig. 21a. This may include the use of shorter striker bars,
Although no published results from dynamic test- and changing the pulse shaper. Thinner samples
ing of Berea (Ohio) Sandstone could be found, results with a lower L:D ratio may also be required to
from another sandstone were found and included here assist in achieving dynamic equilibrium.
for the sake of providing a reference to what may be
expected from a sandstone type of rock. The results of Ultimately, the results from the testing are con-
this testing, performed by Liu et al. (2011), are shown firmed by past published research. Additional testing
in Fig. 21b. This testing was carried out using a should be carried out to solidify these findings. These
3.94 in. (100 mm) SHPB, on samples possessing materials tested in this work commonly used in
diameter 3.82 in. (97 mm) and length 1.69 in. construction and manufacturing and definition of
(43 mm). Note that this is a length to diameter ratio dynamic properties is highly important to the engi-
of approximately 0.44, compared to 1.0 used in this neering design process. Future testing with this SHPB
research. will explore the properties of geologic materials which
are often subject to dynamic processes in the mining
and civil engineering fields, such as blasting, milling,
etc.

123
Geotech Geol Eng

a Berea (Ohio) Sandstone


18000 124.11
R2 = 0.5054
17000 117.21

16000 110.32

b
15000 103.42

14000 96.53
Failure Strength (psi)

Failure Strength (MPa)


13000 89.63

12000 82.74

11000 75.84

10000 68.95

9000 62.05

8000 55.16

7000 48.26

6000 41.37
1 2 3 4 5 7 10 20 50 100 200 400
Strain Rate (s^-1)

Fig. 21 a Graph showing the plot of strain rate versus failure taken from China Qinling, Taibai Mountain, Shaanxi Province
strength for each sandstone sample, and the overall weak performed by and figure taken from Liu et al. (2011)
positive trend, and b results from dynamic testing of sandstone

Acknowledgements This research was internally funded by Frew DJ, Forrestal MJ, Chen W (2001) A split Hopkinson
the University of Kentucky’s Explosives Research Team. pressure bar technique to determine compressive stress–
strain data for rock materials. Exp Mech 41(1):40–46
Frew DJ, Forrestal MJ, Chen W (2002) Pulse shaping techniques
for testing brittle materials with a split Hopkinson pressure
References bar. Exp Mech 42(1):93–106
Gama BA, Lopatnikov SL, Gillespie JW Jr (2004) Hopkinson
ASM International (1990) Metals handbook, volume 2: prop- bar experimental technique: a critical review. Appl Mech
erties and selection: nonferrous alloys and special-purpose Rev 57(4):223–250
materials, 10 edn. ASM International Gong JC, Malvern LE, Jenkins DA (1990) Dispersion investi-
Broch E (1983) Estimation of strength anisotropy using the gation in the split Hopkinson pressure bar. J Eng Mater
point-load test. Int J Rock Mech Min Sci Geomech Abstr Technol 112(3):309–314
20(4):181–187 Gray GT III (2000) Classic split Hopkinson pressure bar testing.
Chen W, Zhang B, Forrestal MJ (1999) A split Hopkinson bar ASM Handb Mech Test Eval 8:462–476
technique for low-impedance materials. Exp Mech Hill JR (2017) Indiana geological and water survey, ‘‘Indiana
39(2):81–85 limestone’’. https://igs.indiana.edu/MineralResources/
Churcher PL, French PR, Shaw JC, Schramm LL (1991) Rock limestone.cfm. Accessed 11 Mar 2018
properties of Berea sandstone, Baker dolomite, and Indiana Holt JM (1996) Structural alloys handbook, Technical Ed; Ho
limestone. Soc Pet Eng. https://doi.org/10.2118/21044-ms CY (ed) CINDAS/Purdue University, West Lafayette, IN
Davies EDH, Hunter SC (1963) The dynamic compression Kaiser MA (1998) Advancements in the split Hopkinson bar
testing of solids by the method of the split Hopkinson test. Diss. Virginia Tech
pressure bar. J Mech Phys Solids 11(3):155–179 Liu S et al (2011) SHPB experimental study of sericite-quartz
Forster IR (1983) The influence of core sample geometry on the schist and sandstone. Chin J Rock Mech Eng
axial point-load test. Int J Rock Mech Min Sci Geomech 30(9):1864–1871
Abstr 20(6):291–295

123
Geotech Geol Eng

Lu YB, Li QM (2010) Appraisal of pulse-shaping technique in Silva J, Lamont R (2017) Dispersion signal analysis in a Split-
split Hopkinson pressure bar tests for brittle materials. Int J Hopkinson pressure bar at the University of Kentucky.
Prot Struct 1(3):363–390 Fragblast Int J Blast Fragm 11(1):7–22
Manes A et al (2011) Analysis of strain rate behavior of an Al Subhash G, Ravichandran G (2000) Split-Hopkinson pressure
6061 T6 alloy. Proc Eng 10:3477–3482 bar testing of ceramics. ASM International, Materials Park,
Petrov Y, Selyutina N (2015) Scale and size effects in dynamic OH, pp 497–504
fracture of concretes and rocks. In: EPJ web of confer- Wu XJ, Gorham DA (1997) Stress equilibrium in the split
ences, vol 94. EDP Sciences Hopkinson pressure bar test. J Phys IV 7(C3):C3–91
Ravichandran G, Subhash G (1994) Critical appraisal of limiting Zhang QB, Zhao J (2014) A review of dynamic experimental
strain rates for compression testing of ceramics in a split techniques and mechanical behaviour of rock materials.
Hopkinson pressure bar. J Am Ceram Soc 77(1):263–267 Rock Mech Rock Eng 47(4):1411–1478
Shankland TJ, Johnson PA, Hopson TM (1993) Elastic wave
attenuation and velocity of Berea sandstone measured in
the frequency domain. Geophys Res Lett 20(5):391–394

123

You might also like