Escolar Documentos
Profissional Documentos
Cultura Documentos
Title no.92-M5
EXPERIMENTAL PROGRAM
Concrete specimens as shown in Fig. 1 were first tested in
a standard material test machine at strain rates of
approximately 10–7/sec to 10–3/sec. The lowest strain rate
of 10–7/sec was used as a basis for normalizing the higher
strain-rate data. This low strain-rate data will be referred
to in this study as static data.
Higher strain-rate tests from 1/sec to 103/sec were
conducted in the SHPB. The SHPB consists of four basic
parts: a striker bar, incident bar, transmitter bar, and spec-
imen. A schematic of the SHPB is shown in Fig. 3. In the
SHPB compressive test mode, the striker bar is impacted
against the incident bar, which induces a compressive
stress pulse in each bar. The striker and incident bar
remain in contact until the reflected wave from the free
end of the striker reaches the interface between the striker
and incident bars. Therefore, the length of the loading
pulse is twice the transit time of the striker bar. The finite-
length stress pulse of the incident bar impinges on the
specimen, sandwiched between the incident and
transmitter bars. Part of the pulse is reflected from the
specimen and part of the pulse is trans- mitted through
the specimen. The reflected and transmitted pulses are
recorded, respectively, at the strain gage positions on the
incident and transmitter bars. The strain gage signals are
recorded on a digital oscilloscope. It has been shown 2
from the general SHPB assumption of compressive
uniform stress along the specimen length that the integral
of the reflected pulse is proportional to strain in the
specimen, the transmitted pulse is proportional to stress
in the specimen, and strain rate is proportional to the
reflected pulse. Using this information, a dynamic stress-
strain curve may be generated for any given strain rate.
Operation of the SHPB in the direct tension mode
differs slightly from operation in the compressive mode
in that a tensile stress pulse impinges on the specimen, but
the funda- mental analysis is the same. For the splitting
tensile tests, the SHPB is operated in the compressive
mode. However, the specimen is rotated 90 deg so that the
load is applied diamet- rically along the length of the
specimen (Fig. 2). The split-
lation of the splitting tensile experiments have been Table 3—Summary of SHPB square-notch results
presented in References 6, 7, and 10. Load Incident Transmitted Dynamic Loading
case stress σI, stress στ, tensile stress rate σ· , Strain rate
no. MPa MPa ftn, MPa MPa/sec ε· , sec–1
Direct tension tests
The direct tension test has seldom been used to evaluate 1 26.5 4.5 9.31 6.01 × 106 4.9
the tensile strength of concrete. This is because of the diffi- 2 67.0 4.0 7.59 1.48 × 10 6 5.3
culties of holding the specimens to achieve axial tension and 3 75.0 4.1 7.93 2.08 × 106 5.8
uncertainties of secondary stresses induced by the holding Note: 1 MPa =145 psi.
devices. Recently, however, direct tension tests using the
SHPB have been successfully conducted.4 Two types of
tensile specimens were tested: a square-notch specimen [Fig. and
1(b)] and saddle-notch specimen [Fig. 1(c)]. Speci- ·
σ
mens were 51 mm in diameter and length. All specimens ε· ----- (4)
were cemented to the ends of the SHPB with a nonepoxy = Es
concrete cement. The bar and specimen surfaces were cleaned
in a manner similar to that used for surface cleaning before the where τ is the time lag between the start of the transmitted
placement of foil-electrical resistance strain gages. stress wave and maximum transmitted stress, and Es is the
Numerical analyses were performed on SHPB direct static modulus of elasticity of the specimen material. The
tension tests conducted for three different loading condi- static modulus of elasticity was used because only limited,
tions. A stress-versus-time history for a typical test is illus- highly scattered dynamic data are available. Based on other
trated in Fig. 10. In these tests, it is assumed that the dynamic experimental data accumulated by the authors,4,7 the
tensile strength at the notch ftn is proportional to the trans- modulus of elasticity of concrete appears to be relatively
mitted stress σT through the expression strain-rate independent. A summary of the results obtained
from the SHPB square-notch tests are presented in Table 3.
ftn = σTAr (1) To provide an accurate numerical simulation of the SHPB
direct tension tests, a detailed FEM model of the specimens
in which and portions of the incident and transmitter bars was
constructed. A partial illustration of the axisymmetric FEM
2 model for the square-notch analyses is presented in Fig. 11(a).
D
b
Ar = ------2 -- (2) The incident and transmitter bars are each comprised of 1633
D bn nine-node, axisymmetric finite elements, and the specimen
is comprised of 476 elements. The load was applied
where Ar is the area ratio, Db is the diameter of the Hopkinson uniformly at the free end of the incident bar. The particular
bar, and Dn is the diameter of the specimen at the notch. FEM topology employed in the analyses was based on wave
Additionally, the loading rate σ· and strain rate ε· in the propagation considerations.6,8,12
specimen can be estimated from the expressions11 The cracking sequence for a direct tension square-notch
specimen for a strain rate of 5.3/sec (load case 2) is presented
f tn
σ· in Fig. 12. The nonlinear FEM analyses indicate cracking at
both the center notch and incident end of the specimen.
44 ACI Materials Journal/January-February 1995
= ----- (3) However, the predicted mode of failure is dependent on the
τ
strain rate. For a low strain rate (load case 1), failure occurs
at the notch; for an intermediate strain rate (Fig. 12), failure
occurs simultaneously at the notch and incident ends of the
specimen; and for a high strain rate (load case 3), failure
occurs at the incident end of the specimen.12 Similar
behavior patterns were exhibited in the saddle-notch experi-
ments and numerical simulations.
In both the experimental and numerical analyses, the direct
tension failures appear to occur in the rise-time portion of the
able with loading rates. These two solutions are also illus-
trated in Fig. 4 and compare very favorably with the
experimental data.
It has been suggested that the strain-rate sensitivity of
Fig. 13—Time history predicted by FEM analysis for longi- concrete may be attributed to the presence of water in the
tudinal stress σz in square-notch specimen for load case 2 cured concrete. The effect of water content on concrete
(see transmitted trace of Fig. 10) (1 MPa = 145 psi).
strength at high strain rate was experimentally determined by
first establishing concrete drying curves (Fig. 5) and then
testing concrete specimens with various water contents in the
SHPB. Results of the dry concrete tests are shown in Fig. 6.
approximately 75 percent of the specimen has been crushed. Results for the completely wet concrete, with the same mix
At time t = 275 μsec [Fig. 14(d)], most of the specimen has as the dry concrete, are shown in Fig. 7, and results for the
been crushed (approximately 85 percent) and failure occurs. half-dry concrete specimens are presented in Fig. 8. It
This simulation suggests that the specimen is pulverized at appears from Fig. 7 and 8 that the wet and partially wet
failure, consistent with the observed experimental results for concretes show a higher strain-rate sensitivity than dry
this high strain rate. concrete of the same mix. However, if one compares the
static strength of the wet concrete to the dry concrete (Table 2),
DISCUSSION OF RESULTS the wet concrete has less strength than the dry concrete. If the
The sensitivity of concrete strength to strain rate is clearly findings of Reinhardt were applied here,15 i.e., as concrete
illustrated in Fig. 4. However, the threshold or critical strain strength decreases, low strain-rate sensitivity increases, one
rate at which concrete exhibits significant strength increase might expect a higher strain-rate sensitivity for the wet
is different for tension and compression, approximately 5/sec
concrete. However, it is assumed that the Reinhardt data15
and 60/sec, respectively. Also, the tension curve exhibits
are associated with concrete quality of different concrete
much higher strain-rate sensitivity than the compression
mixes and would not necessarily apply here for different
data. The slope of the tension curve beyond a strain rate of
strengths between wet and dry specimens of the same mix.
2.5/sec may be described by the equation
With this in mind, increased rate-sensitivity and strength of
the wet specimens at the upper end of the low strain-rate
ftd = K ε· l/n (5) range are an effect due to the presence of excess moisture.
Effect of excess moisture is shown by the dashed curve of
in which ftd is the dynamic tensile strength, K is a proportion- Fig. 7. However, the slope of the strain-rate curve above a
ality constant, ε·is the strain rate, and l/n is the slope of strain rate of 50/sec is not expected to change. This is in good
log10(ftd) versus log10( ε·). For the tensile data presented in agreement with experimental data. It has recently been
Fig. 4, the value of n is approximately 2.5. Using fracture shown experimentally16 that dry concrete shows no rate-
mechanics of brittle materials, Grady and Lipkin13 showed sensitivity up to strain rates of approximately 1.0/sec for
theoretically the value of n to be 3.0. direct tension, but wet concrete for the same test region
Using the uniformly distributed array of penny-shaped shows rather high rate-sensitivity. The same general conclu-
cracks, Weerheijm and Reinhardt14 calculated concrete sion may be drawn here for direct compression of concrete.
tensile strength for high-quality concrete (compressive Crack patterns generated by numerical analysis of the
strength of 50 MPa). Analytical solutions were performed splitting tensile test6,7,10 indicate that fracture does not
using a constant fracture energy and a fracture energy vari- initiate on the mid-diameter of the specimen for dynamic
ACKNOWLEDGMENTS
loading. This prediction agrees precisely with experimental The experimental work was conducted at Wright Laboratory Airbase
results. Cracking begins between the specimen center and Systems Branch, Tyndall Air Force Base, Florida. The numerical analysis
incident side of the specimen. The growth of this initial was sponsored by a research contract from Wright Laboratory Airbase Sys-
surface crack in the numerical analysis is much faster than in tems Branch, Contract No. F08635-90-C-0120. Computational resources
were provided by the Alabama Supercomputer Network.
the experiment. This is due to the fact that the crack-tip
velocity in the experiment is much less than the acoustic
REFERENCES
velocity in concrete, as shown in Fig. 15. This implies that, 1. Kolsky, H., “Investigation of the Mechanical Properties of Materials at
in the numerical analysis, the crack will travel at the same High Strain-Rates of Loading,” Proceedings of the Physical Society,
speed as the stress wave because there is no provision in the Section B, V. 62, 1949, pp. 676-700.
material model to account for a lower crack speed. The 2. Nicholas, T., Impact Dynamics, J. Wiley and Sons, New York, 1982,
pp. 277-332.
results of the splitting tensile numerical analyses are in good
3. Jerome, E. L., “Analysis of a Proposed Six Inch Diameter Split Hopkinson
agreement with the experimental results, except that the frac- Pressure Bar,” PhD dissertation, University of Florida, Gainesville.
ture occurs sooner in the numerical simulation. 4. Ross, C. A., “Split Hopkinson Pressure Bar Tests,” ESL-TR-88-82 HQ
In the numerical simulation of the direct tension tests, AF Engineering and Services Center, Tyndall AFB, Florida, Mar. 1989.
5. ADINA, “Finite Element Computer Program for Automatic Dynamic
cracking occurred at both the notch and the incident end of
Incremental Nonlinear Analysis,” Report ARD 84-1, ADINA R&D Inc.,
the specimen. Depending on strain rate, failure could occur Watertown, Massachusetts, Dec. 1984.
at either or both locations. Identical behavior was observed 6. Tedesco, J. W., “Numerical Analysis of Dynamic Splitting-Tensile
in the experiment. Failure occurred at either location for and Direct Tension Tests,” ESL-TR-89-45 HQ AF Engineering and Services
similar low-rate loadings, and multiple failures (both notch Center, Tyndall AFB, FL, Sept. 1990.
7. Tedesco, J. W.; Ross, C. A.; and Kuennen, S. T., “Experimental and
and incident end locations) occurred at high strain rates. This Numerical Analysis of High Strain-Rate Splitting-Tensile Tests,” ACI
multiple fracture phenomenon was the primary factor Materials Journal, V. 90, No. 2, Mar.-Apr. 1993, pp. 162-169.
prompting the splitting tensile test program in the SHPB to 8. Tedesco, J. W.; Hughes, M. L.; and O'Neil, B. P. “Numerical Analysis
assess the dynamic tensile strength of concrete. of Dynamic Direct Tension and Direct Compression Tests,” ESL-TR-91-41
HQ AF Civil Engineering Support Agency, AFCESA/RACS Tyndall AFB,
In the SHPB direct tension tests of concrete, failure Florida, Feb. 1992.
occurred in the rising portion of the loading curve, as shown 9. Meyer, C., and Bathe, K. J., “Nonlinear Analysis of Reinforced
in Fig. 10. Following the fracture, an unloading stress wave Concrete Structures in Practice,” Journal of the Structural Division, ASCE,
is “trapped” in the broken specimen, which is still attached V. 108, 1987, pp. 1605-1622.
10. Tedesco, J. W.; Ross, C. A.; and Brunair, R. M., “Numerical Anal-
to the transmitter bar. Due to the mismatch of characteristic ysis of Dynamic Split Cylinder Tests,” Computers and Structures, V. 32, No.
impedance of the steel bar and concrete, the reflected pulse 3, 1989, pp. 609-624.
at the bar-concrete interface is approximately the same 11. Malvern, L. E., and Ross, C. A., Dynamic Response of Concrete Struc-
magnitude and bears the same sign as the incident pulse. This tures, First Annual Technical Report, Air Force Office for Scientific Research,
Bolling AFB, Washington, D.C., Contract F49620-83-K007, 1984.
“trapped” pulse then produces an oscillating signal in the
12. Tedesco, J. W.; Ross, C. A.; McGill, P. B.; and O'Neil, B. P.,
transmitter bar. The numerical analysis results illustrate this “Numerical Analysis of High Strain-Rate Concrete Direct Tension Tests,”
same oscillation in the specimen (Fig. 13). Computers and Structures, V. 40, No. 2, 1991, pp. 313-327.
The results of the FEM simulations of the direct compres- 13. Grady, D. E., and Lipkin, J., “Criteria for Impulsive Rock Fracture,”
Geophysical Research Letters, V. 7, No. 4, Apr. 1980, pp. 255-258.
sion experiments predicted failures consistent with those
14. Weerheijm, J., and Reinhardt, H. W. “Modeling of Concrete Fracture
observed in the experiments, i.e., an increasing number of under Dynamic Tensile Loading,” Fracture of Concrete and Rock: Recent
smaller fractured pieces for increasing strain rate. These Developments, S. D. Shah et al., eds., Elsivier Applied Science, London,
simulations also verified the SHPB assumption of uniform 1989, pp. 721-728.
stress in the specimen.8 15. Reinhardt, H. W., “Dynamic Loading,” Fracture Mechanics of