Simulation of Earth Penetration Shock Using High-Speed Impact into an Engineered Water Target, Part 1
N.T. Davie, K.E. Metzinger Sandia National Laboratories, PO Box 5800, Albuquerque, NM 87185
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy??s National Nuclear Security Administration under contract
ABSTRACT Earth and hard-target penetrator shock environments can be simulated with a new test technique developed to produce highg, long-duration acceleration pulses, with an associated velocity change that is much higher than available with traditional shock testing methods. This technique utilizes a controlled high-speed impact with an engineered water target and is implemented on Sandia??s 10,000 ft rocket sled track. Momentum exchange with the water target produces controlled and repeatable decelerating forces on a rocket sled containing a test item. After engagement with the water target, the sled continues at reduced speed and is gradually slowed to rest on the track for post-test recovery of the test item. The test sled is reusable, which lowers test cost and reduces turn-around time. The new capability allows laboratory-like testing to a prescribed acceleration-time history, which makes this technique suitable for component qualification and model validation. This paper, part 1 in the series, includes a review of the test concept along with initial experimental results. Numerical methods are described, which model the sled/water interaction as well as the structural response of the sled. The numerical/experimental process resulted in a successful sled and target design capable of subjecting large payloads (up to 1200 lb) to high acceleration, long duration shocks. Experimental results along with numerical comparisons are also described. INTRODUCTION The work described in this report was motivated by the need for a low-cost test capability for simulating penetration impact shock on weapon components and large subsystems. Although this project was stimulated by specific test requirements, the goals of the project were established to serve a broad customer base, including penetrator weapons, impact sensors, aerospace systems, and in general any system that requires high-g, long duration shock testing. The specific goals for the new shock test capability were: rigid-body shock amplitude up to 3000 g??s, velocity change up to 1000 ft/sec, and payload weight up to 1200 lb. A high-speed earth penetrator impact produces a severe mechanical shock environment
on the penetrator structure and its functional components. This environment is characterized by high-g deceleration, commonly 1000 g??s or more, and persisting for 5 to 30 milliseconds. Assuring that the penetrator structure and its contents survive and function during this severe environment is one of the primary design obstacles, and is overcome with creative design, analysis, and testing. The mechanical design process begins with an estimate of the expected environments, based on simple analyses or knowledge of similar systems. Components and structures are designed or selected to survive this environment. Experimental qualification starts with low-risk component tests, and progresses to increasingly complex higher-risk tests, culminating with full-system tests. It is impractical to qualify components with full-system testing due to their inherently high cost, often well over $1 million per test. In addition, full-system hardware is not available until later in the design process, and full-system tests cannot produce quantifiable test margin (i.e. they generate the nominal environment). Computational tools for structural analysis have advanced dramatically, and they can be used to plan smarter tests, but they do not eliminate qualification testing. At the very least, testing is necessary to validate models, and to determine if all of the pertinent physical phenomena have been included in the model. The challenge for the lower-cost, lower-risk component and subsystem test is that it must adequately simulate the shock produced by the actual impact, and with quantifiable margin (over-test). Model validation is an increasingly important part of any development project, as tight budgets and improved computational capabilities supplant the most costly parts of a test program. These model validation experiments often require extensive instrumentation and state-of-the-art measurements that are best served with controlled and repeatable test techniques. Table 1 lists several test alternatives ranging from low to high cost, including relative merits and capabilities. The commercial drop test machine would be an ideal component or subsystem test method, due to its low cost and ability to produce quantifiable over-test, all with low-cost reusable test fixtures. The ability to control and repeat the shock environment makes this method well suited for model validation. However, its low velocity change (<200 ft/sec) is inadequate for most penetrator environments. Other methods shown in Table 1, while having adequate velocity change, require expensive targets (typically
concrete) and expendable penetrator case structure. Concrete targets may have material properties that significantly deviate from design properties, and they may be difficult to design so as to produce a quantifiable over-test. Hence, there is motivation to develop a high-speed test method that is relatively low cost and produces a controlled and repeatable shock environment for testing penetrator
components and systems.
Table 1. Relative merits and capabilities of selected test methods. Test Method Relative Cost Can peak g??s be attained? Yes Yes Yes Yes Yes Velocity change (ft/sec) <150 <2000 <1000 1000 or more 1000 or more Produces quantified test margin? yes no no no no Reusable test fixtures yes no no some some Requires expensive ??hard?? target no yes yes yes yes Requires expendable penetrator case no yes yes yes yes High launch g??s no yes no no no Max. test item size medium small large large large
drop test machine airgun or Davis gun Aerial cable Conventional rocket sled Full-system
Very low low moderate moderate Very high
TEST CONCEPT To meet these requirements, with a new test method has been developed using the Sandia National Laboratories?? 10,000 ft. rocket sled track . A sled containing the test item and on-board instrumentation is propelled into an engineered water target that produces the desired deceleration (by means of momentum transfer). Following impact the sled continues along the track at reduced speed until it is gradually brought to rest where the test item and data acquisition package are easily recovered. All test hardware is reusable except for the spent rocket motors and low-cost water target. This general test concept is depicted in Figure 1. With the proper water target design, an arbitrary acceleration-time profile can be produced within relatively broad limitations. This method does not duplicate the interaction of a real target with a penetrating device. However, once the desired acceleration profile is known, it can be produced repeatedly for component testing and model validation. The unique features of this new test technique are: 1) direct impact into a low-cost water target to produce a controlled and repeatable high g deceleration profile, with associated high velocity change, and 2) impact occurring while the sled is on the track (as opposed to the traditional impact off the end of the track), allowing controlled posttest deceleration and recovery with a reusable test sled. 1st stage sled sled track water target in polystyrene box test item sled (optional)
post-test sled recovery
Figure 1. Test Concept
The momentum exchange principle and governing equation of motion are shown in Figure 2 , where a flat-faced sled strikes a water target at speed V, with water being ejected at a right angle relative to the sled motion. This equation can be used to design the cross-sectional area of the water as a function of target length. The concept of using water to stop rocket sleds is not new, where water in a trough between the rails is commonly used to gradually stop first-stage sleds at deceleration amplitudes below 50 g??s. However, the new shock test
concept requires deceleration almost two orders of magnitude greater, with associated forces up to 8 million pounds for the maximum payload weight! In order to survive these large forces, the sled must be a substantial structure, and stiff enough so that the dynamic response of the sled does not cause undesirable
superposition with the input shock pulse. In addition, the center of mass of the target must align with the projected center of mass of the sled, to prevent large eccentric loads that would rip the sled off of the track. This alignment is readily facilitated by using a polystyrene foam container for the water. The mass of the polystyrene foam is negligible compared to the water.
sled stationary water target with area A, density ?Ñ
?ÑAv F=?Ñ 2
Figure 2. Momentum exchange principle and equation of motion.
PROJECT OUTLINE AND DESIGN PHILOSOPHY Development of the new test concept into a mature test method required a balanced approach utilizing both testing and analysis. The complementary interaction between analysis and testing ultimately reduced cost and risk, and shortened development time, compared to alternate approaches that rely too heavily on either analysis or test. Experiments proceeded from relatively simple to complex, and provided learning opportunities to improve experimental techniques. Computational models and techniques were developed in parallel with experiments, giving confidence to proceed with, or modify planned experiments. An incremental approach was taken at each experimental phase of the project, where the shock amplitude would be increased from roughly half to full level in two or three steps. Omission of lower amplitude experiments, significantly increases risk, and may result in a test where the only information learned is ??it didn??t work??. Results of each of the experiments provided validation information for the corresponding numerical model, and allowed an assessment of whether all the important physical phenomena had been included in the model. Agreement between analysis and test throughout the process, yielded increasing confidence in the numerical techniques, which ultimately allowed lower design margins and a reduction in sled weight. Using this development philosophy, the following project outline was implemented. 1. 2. Preliminary analysis. Confirmation of momentum exchange equation allowed initial experiments to proceed. Initial (or learning) experiments (half level and full level). Results validated model of sled and water interaction forces. Water forces on the sled track were significant but had not been included in the model. Results led to improvements in target construction. Extensive analysis/design phase. Validated model allowed iteration/optimization of sled shape to direct ejected water away from
the track, and provided an estimate of forces on the rails. This was followed by detailed analyses of the sled structure to minimize sled weight while providing a design capable of surviving the high-g impact. Validation experiments. Again, an incremental approach was used, with the first test at approximately half amplitude. Results provided increased confidence in the model, and proved the water ejection concept. Tests of customer payload. Two tests of penetrator subsystems were conducted at the required shock amplitude. Confidence in the numerical model gained in step 4 allowed a significant increase in the original payload weight requirement to be accommodated. The model was used to direct minor design changes, and accept lower design margin.
INITIAL EXPERIMENTS AND RESULTS Prior to conducting the initial tests, a computational study was performed to determine if the momentum exchange principle could be modeled with existing computer codes and techniques. For this and subsequent analyses, the water target was modeled with a gridless Lagrangian Smooth Particle Hydrodynamic (SPH) technique  implemented in PRONTO3D , which is a general-purpose, transient dynamic, nonlinear finite element program. The geometry chosen for this initial analysis consisted of a 7 in x 7 in x 30 ft long water column that was struck by a 1200 lb aluminum sled (~23 in cube) at an impact speed of 2200 ft/sec. Results of the analysis , compared favorably with the momentum exchange equation, and provided confidence to proceed with initial tests. The first test was designed to produce approximately 1100 g??s for 20 milliseconds. The sled, shown in Figure 3, consisted of thick aluminum plates attached to each end of a tubular mock payload. The 1430 lb sled was propelled with eight Super Zuni rocket motors to an impact speed of 1300 ft/sec, with the 21?? long water target, shown in Figure 4. Results, shown in Figure 5, show higher than expected deceleration during the first 10 milliseconds. This was caused by water permeating the unlined polystyrene target, which effectively allowed more water area than designed. SPH analysis, using the as-designed target
dimensions, shows closer agreement with the intended shock amplitude. Some outward displacement of the track also occurred, due to the force of the ejected water. Significantly higher deceleration was desired for the second test, which would produce even larger forces on the track, possibly resulting in failure. SPH analyses indicated that the water could be partially directed away from the track by placing raised diversion structure on the face of the sled, which was implemented for the second test. In addition, improvements were made to the target, including lining the interior with water-resistant mastic, which reduced water permeation. Two-stage rocket propulsion was required to
attain the desired 1964 ft/sec impact speed of the 1610 lb sled. The first stage pusher sled was propelled with 25 Super Zuni rocket motors and the test item sled with 8 Super Zuni rocket motors, as before. Measured sled deceleration, shown in Figure 6, compares favorably with the designed shock amplitude. However, the water diversion structure did not adequately protect the track, which received significant damage. Even though this test was not completely successful, it provided data for model validation, and gave direction for subsequent analyses, which would focus on water diversion, and track protection.
Figure 3. Test 1 sled, prior to motor installation.
Figure 4. Water Target for Test 1.
Acceleration in G's
idealized design pulse SPH analysis
Acceleration in G's
Time in Seconds
Figure 5. Test 1 results. Figure 6.
Time in Seconds
Test 2 results, showing idealized design pulse along with measured data.
WEDGE SLED DESIGN The results of the first two tests showed that the water impingement forces on the sled track were greater than originally expected. This result stimulated a significant design and analysis effort to arrive at a robust sled design that would also divert water away from the track. The final sled design is presented here,
while computational details and results are described in the following sections. Figure 7 shows the final sled and water target design. The wedge-shaped sled and two-part target allowed water to be directed primarily to the sides and away from the track. Although this new design minimized forces on the track, it was significantly heavier and more complex that the original design. The sled was constructed with 7010-T652 highstrength aluminum alloy plate arranged in 7 layers, which were held together with 24 high-strength steel rods (? 1.25??) made of 4340 steel. Shear connection between the layers was achieved with aluminum splines that fit in grooves in the plates. The use of plate allowed an internal cavity for weight reduction, and provided better material properties than either a casting or forging. A 0.032 thick stainless steel ??skin?? was bonded to the forward surface to prevent the possibility of water entering at the plate interfaces. Overall dimensions of the wedge sled were 40?? wide and 34?? tall. The two-part target was 24?? tall (water depth), with a maximum water width of 6?? for each side. The center-to-center spacing of the targets was 21.5??, and the target was positioned so that the center of each side would intercept the corresponding slanted face of the sled. The problem of water permeation was solved by using a 4 mil thick polyethylene liner. The target shown, had a 24?? long main (shock pulseshaping) target, followed by about 80?? of low-g braking targets. The sled was propelled with 4 Javelin rocket motors, and 2 Marc 46 motors, to achieve approximately 2000 ft/sec impact speeds. The 48?? long cylindrical payload was centered on the back face of the wedge.
Figure 7. Wedge sled and target
FINITE ELEMENT MODELS Scores of finite element models have been created in support of this project; two of them will be described in detail. Figure 8 shows a monolithic sled face, a water target, and a rail in red, blue, and green, respectively. This type of model was used to investigate the efficacy of various designs intended to channel the water away from the rails. The sled face and the rail are modeled with traditional hexahedral finite elements. The water is modeled with SPH (smoothed particle hydrodynamic) elements. SPH is a gridless Lagrangian technique which has been coupled to the PRONTO3D transient dynamics code. The resulting finite element code is capable of modeling the structural response of systems that experience very large deformations while avoiding mesh distortion problems. Note that the entire sled is not modeled nor is the low-density polystyrene foam which holds the water in place prior to impact. The weights of the sled components which are not modeled are added to the sled face. Only one rail was modeled to allow an independent, momentum-based check of the lateral loads the water imparts on the rail. A line of nodes on the bottom surface of the rail is fixed in the lateral and vertical direction (every four
feet) to represent the restraint provided by the track clips. The sled is constrained in the both the vertical and lateral directions to represent the restraint provided by the sled shoes. After numerous iterations, a wedge-shaped sled face with a split water target was selected. More detailed models were subsequently created to determine the internal plate details and the number and size of the rods. Figure 9 shows the finite element mesh used to predict the rod and plate stresses. Numerous chamfers and radii were included in this model to improve convergence in the static (bolt preload) analysis and accuracy in the corresponding dynamic (water impact) analysis. As shown in Figure 9, the individual plates (red and green), rods (yellow) and splines (light blue) of the wedge were modeled. In the actual hardware, the 10 rods in the vicinity of the shoes terminate in blind holes in the bottom
plate. The other 14 rods are secured with a nut on the bottom of the wedge. All of the rods are secured with a nut at the top of the wedge. Only the ends the rods are threaded; the middle sections of the rods are smooth. The finite element model includes the top nuts, but not the bottom nuts. Instead, the bottom ends of all 24 rods extend through the bottom plate to allow the bolt preload to be incorporated into the analysis. No threads are included in the model. Note that only the sled face is modeled. The extra weight associated with the payload, support tube, rocket motor cases, back plate and sled shoes is added to the plates in the sled face. Each plate is assigned a different density to reflect the vertical weight distribution of the items that are not modeled. Tables 2 and 3 show the material properties [4, 5, 6] used in this study.
Table 2: Material Properties - Solids Part Monolithic Sled Rail Composite Sled Splines Composite Sled Plates Composite Sled Rods Material Aluminum Steel Aluminum Aluminum Steel Elastic Modulus (psi) 6 10x10 6 30x10 6 10x10 6 10x10 6 30x10 Poisson??s Ratio 0.3 0.3 0.3 0.3 0.3 Density 2 4 (lbfs /in ) -4 6.565x10 -4 7.76x10 -4 2.59x10 varies -4 7.324x10
Table 3: Material Properties ?C Water (Mie-Grueisen Equation of State) Density 2 4 (lbfs /in ) -5 9.585x10 Pressure Cutoff (psi) -1 C0 (in/s) 58267 S 1.921 Gamma 1
The metallic materials are assumed to remain elastic in the analyses. However, the elastic limits of the materials must be known in order to establish design margins. Accordingly, tensile tests were conducted on specimens machined from the plates and a few spare rods were also loaded to failure. The Al 7010-T7651 stock used for the plates has a yield strength of 74 ksi in the transverse direction. The 1.25 inch diameter rods were made of 4340 steel and have a proof load of 160, 000 lb. (The proof load is defined to be the lowest load at which permanent deformation occurs.)
Figure 8: Finite Element Mesh (Monolithic Sled, Water, and Rail)
Figure 9: Finite Element Mesh (Composite Sled)
TRACK LOAD ANALYSES (MONOLITHIC SLED MODEL) Figure 10 shows the water spray pattern predicted by a PRONTO3D/SPH simulation for the baseline water target 10 ms after impact. Most of the water turns approximately 90 degrees and misses the rails after it leaves the sled face. However, a small portion of the water strikes the rails or moves forward between them. Figure 11 shows the vertical rail loads for both the baseline and a larger water target. The baseline water target is 24 inches tall with a maximum width of 6 inches for each half. The larger water target is 26 inches deep with a maximum width of 6.3 inches for each half. The same impact speed was used for both targets. The loads were obtained by summing the reaction forces at the constrained nodes. The excitation of a rail bending mode causes the significant difference between the 200 and 500 Hz curves for the larger water target. This type of resonance probably contributed to the track damage observed in the flat-faced sled tests. There??s a big incentive to use larger targets because they could allow some high-level tests to be done at a lower impact velocity with a single stage. Although the difference between these two water targets is not great, the larger water target leads to a big increase in the rail loads. In addition, the actual increase in the acceleration (~5% in the plateau region) is smaller than would be expected because some of the water is not turned as completely with the larger target. These analyses indicate that there is substantial risk and a only modest reward for using water targets larger than the baseline design.
Figure 10. Baseline Water Target (front view), 10 ms After Impact
Figure 11. Rail Loads, Baseline and Larger Water Targets
COMPOSITE SLED ANALYSES The static bolt preload is accomplished with a sequence of JAS3D  simulations using the conjugate gradient solution technique. With the bottom surface of the wedge constrained, the bolt preload is applied to the bottom of each rod. The rods are free to slide through the holes in the plates as they elongate and the plates are compressed. After a converged solution is obtained, the portions of the rods which are inside the bottom plate are tied to this plate. The external loads are subsequently released. The JAS3D results for the preloaded wedge are then mapped into a model that includes the desired water target and the wedge is given its initial velocity. The resulting stress state can now be imported into PRONTO3D/SPH. Figure 12 shows the predicted water spray pattern 10 ms after impact. Although the main purpose of the full-level tests was to test weapon hardware, some instrumentation was also included to measure the sled response. Figure 13 shows the predicted and measured acceleration-time histories for the sled (note that the amplitude has been normalized by the specific peak amplitude of the test requirement which was in excess of 2000 g??s).
The close agreement between these curves validates the technique used to simulate the water loading. Figure 14 shows the locations of the strain gages attached to the wedge. The gages on the aluminum plate measure primarily bending strain in a horizontal direction. The gages on the steel rod measure primarily axial strain. Figure 15 shows that the predicted and measured responses for gages 1, 2 and 3 agree fairly well, although there was a problem with gage 1 sometime after the peak load,. Of course, the biggest concerns are what are the highest plate stresses and where do they occur. This takes some time to determine because the predicted plate stresses are rather noisy due to water-to-plate and plate-to-plate contact. In addition, the model is fairly large (~1.2 million elements) which restricts the content and frequency of the global output. For the current design, the von Mises stress at strain gage 2 is as high as it is at other critical locations. In addition, the peak von Mises stress at this location is dominated by a single component - the lateral stress. Thus, the measured peak of 4000 microstrain corresponds to a peak von Mises stress of 40 ksi, which is well below the transverse direction yield strength of 74 ksi.
Figure 12. Composite Sled, 10 ms After Impact
Figure 13. Axial Acceleration (1 kHz)
Figure 14. Strain Gage Location Figure 16 shows the predicted and measured values for the average strain in the rod. The response of the model to vertical expansion may be a bit slow, since the plates incorporate the extra weight of items not included in the model. Nonetheless, the agreement is reasonable except for the localized peak in the test data. A careful review of the data and the gages indicates that this peak is just a noise spike which has been filtered to the point that it almost looks like real data. Thus, this peak will be ignored. The axial load can be obtained by multiplying the average strain by the elastic modulus and the cross sectional area. The axial load should be constant along the length of the rod if frictional effects are negligible. The next highest peak in the average strain is much more credible. This peak of 3700 microstrain would indicate a peak load of 136,000 lb. The resulting factor of safety against permanent deformation would be slightly less than 1.2, assuming this same load is also present at the threaded ends. Low factors of safety are commonly used for bolted connections, where bolt preloads are typically specified from 60 to 90% of the proof load  for static applications. The nominal preload and maximum measured load for this application were 63 and 85% of the proof load, respectively.
Temporary data loss on ch 1.
Figure 15. Aluminum Plate Strain Gages
transient data anomaly (filtered spike)
Figure 16. Steel Rod Average Strain SUMMARY A new shock test