CP620, Shock Compression of Condensed Matter - 2001 edited by M. D. Furnish, N. N. Thadhani, and Y. Horie © 2002 American Institute of Physics 0-7354-0068-7/02/$ 19.00 MICROSTRUCTURAL MODEL OF IGNITION FOR TIME VARYING LOADING CONDITIONS Richard V. Browning and Richard J. Scammon Los Alamos National Laboratory, PO Box 1663, Los Alamos, NM 87545 Abstract. A micro-mechanical based model of ignition was developed about five years ago based on a simple inter-granular friction model of mechanical dissipation coupled with a fit to extensive direct numerical simulations of the resulting thermally induced decomposition. The chemical model used was the McGuire-Tarver ODTX based model for HMX decomposition. The resulting power law type model has been reasonably successful in predicting threshold conditions for Steven type experiments. The final power law form was obtained by assuming a constant time history for both the pressure and shear strain rate, resulting in time independent loading conditions for the chemical model. Here we propose to extend the model to handle time varying loading conditions. This is done using a linear operator that models reactive heat transfer simulations done for a wide variety of loading conditions. The linear operator is represented by a convolution integral with Prony series kernel form for efficient numerical implementation. To complete the model the same inter-granular friction model used previously is employed. Comparisons are made with results of numerical simulations and experiments. The technique used here is based on the notion of linearizing the reactive heat transfer problem. Although the chemical model involves four reactions and is highly nonlinear, we effectively linearize the problem around ignition conditions with a linear operator fit. We use a simple power law approximation that gives useful accuracy over at least 4 orders of magnitude in time and fluence. A non-dimensional scaling method is used to determine the final form. We believe the techniques used here could also be used with more detailed chemical models and with other types of mechanical dissipation models. INTRODUCTION Models for ignition of reactive materials are needed for safety studies. They need to incorporate an energy dissipation mechanism, heat conduction, and chemical reaction. In this paper we focus on modeling the heat conduction and chemical reaction mechanisms in a concise form. Browning (1) developed a microstructural based model of ignition based on intergranular friction coupled with fits to numerical simulations of thermal ignition based on McGuire-Tarver (2) kinetics and finite element numerical techniques in Chemical Topaz (3). The result was a very simple power law relation between time of ignition, pressure and shear strain rate, however a basic assumption was that the pressure and shear strain rate was constant up to the time of ignition. The simplicity of the model allowed easy implementation and use compared with using direct numerical simulation of sub-scale reactive heat transfer models (4), and produced reasonable agreement with results from certain classes of experiments (5,6,7). In this paper we investigate a technique for dealing with non-constant time history loading by using a linear operator to model the heat conduction and chemistry involved in the ignition process. Effectively, we linearize about the ignition conditions, incorporating the nearly linear heat 987 conduction process and effectively hiding the severe non-linearities of the detailed chemistry. model, and then verify the accuracy by comparisons with the results. THERMO-CHEMICAL MODELING 1.5- 1.071 \ y= 0.776SX2 ^ R = 0.9743 We start with the thermo-chemical process as defined by the McGuire-Tarver kinetics model implemented in Chemical Topaz. By running the code on simple axi-symmetric problems for varying spot sizes we can calculate ignition conditions for varying spot sizes and flux time histories. To transform the computed results into a more compact numerical model we first develop a nondimensional representation of the results following Barenblatt(S). We assume seven dimensional variables, the critical flux cpcr, critical time tcn spot radius /?, heat capacity cp, thermal conductivity k, density p, and a heat of reaction q. This is simplified from the actual system, where we have three reactions in the chemical model each with separate q values, for example. These variables have five dimensional units, so we can choose two independent nondimensional parameters. We choose a flux related parameter 7CV = (pcrcpRlqk and a time related #/ | 0.5.^ >^ 5" -0.5 • ^ ! -1. W 4^X -1.5^ -21 0 1 2 3 A Log10(Scaled Time) Figure 1. Scaled results from original 2-D calculations. A constant flux is applied over spot sizes 0.25 to 250 micron diameter. The linear fit to logarithmic values is equivalent to a power law fit. To express the relation in terms of real time, we substitute the definitions of the non-dimensional parameters back into the integral representation, and after changing the integration variable obtain a representation, t F2(t) = \K(s2t - s2x)sl<p(x)s2dx , parameter 71'2 — tcrh I R where we used thermal 0 diffusivity X^ — kl pcp as a derived quantity. where we have used scaling factors defined by ^(^2) = S^(t) and 7T2=S2t. Figure 1 shows the scaled results of a series of calculations on spot sizes 0.25, 2.5, 25 and 250 |Lim. These are given in terms of scaled fluence 7Tlb — S^O = 3> I pqR . On this log-log plot the straight line fit is equivalent to a power law function and so we take Our basic assumption is that the non-dimensional parameters are related by a linear operator, Fcr = L(^rj(/r 2 )) . We implement the linear operator as a convolution integral eventually using a Prony series representation for the kernel function. Considering KI a function of 7i2 we can write, 7Tlh (jc) = axn and then the kernel becomes n K(x) = x~ (\-ri)la . Given the power law form for the kernel, we can factor out the scaling terms to obtain the thermo-chemical model in real time variables, and then Fcr =L(^(^ 2 )) . This assumption is valid for linear heat transfer problems as the temperature at any point of the body can be obtained from the response of a linear operator acting on the time dependent boundary conditions. In the case of non-linear reactive heat transfer, the ignition condition is not exactly defined by a linear operator, but we fit computational results to the o The critical time is then defined Fcr = F2(tcr) , where Fcr is a fitting constant. 988 by We use a Prony series kernel in the convolution integral to simplify the integration process. The Prony terms are taken as constant ratios, with the ratios selected to cover the range of times involved, and to generate a good fit to the power law function. Fig. 2 shows a comparison of the Prony model results and the original calculations for various spot radii. The scaling gives a good approximation for the various spot sizes, and the slope of the fluence time curve is accurate, however we do miss the time-to-ignition depending on the distance to the calibration point. This is basically the result of the scatter in the original scaled timeto-ignition data. Some of this scatter is inherent in the approximation, however some also results from numerical effects in the Chemical Topaz calculations. excellent. Other more complex problems are in progress, with good results emerging. OVERALL MODEL To create an overall model we use our previous results(l) on inter-granular friction to complete the model. The spot radius and flux are expressed in terms of the macroscopic pressure P and shear strain rate y , as well as the material properties grain radius RG, coefficient of friction /3 and grain elastic compliance C = 2(l — V)/E, where v i s Poisson's ratio and E is Young's modulus. We use, R = 0.721 • 21/6 - RGCEPm = C2Pl/3 2/3 v and 1 = Cf^Y* where C, =1.031J3C~ R- . /s 10- *** F * 0.25, 1- * * X 0.1 • •Original Calculations X Model calibrated at 2.5b V 0.01 • 0.2 X* 0.1 10 1000 100000 10000000 0.2 Time to ignition (microsec) 0.4 0.6 0.8 Fraction of Time On Figure 3. Comparison of Prony model and Chemical Topaz calculations for one-dimensional geometry. The flux is on for an interval, then off, then back on. The total on time is split evenly between the first and second segment. The agreement is very good indicating the ability to deal with widely separated flux pulses. Figure 2. Comparison of Prony model with results from original model. COMPARISON WITH OTHER HISTORIES To verify the linear operator technique we ran a series of problems with non-constant flux conditions. The numerical reactive heat transfer results are compared with the predictions from the linear operator model. We first ran a series of calculations on a one-dimensional problem. In this series the time history was an on-off-on sequence where the total on time was split equally between the first and second heating pulse. The results are compared in Fig. 3 in terms of the reduction of the ignition criterion. In this case the agreement is After combining all these expressions, most of the constants are merged into the fitting constant Fcr, but we retain the explicit dependence of R within the integral, to give the final expression t = J(r - We have tried fitting this with n=0.7768, and the critical value of F taken at several different times. The fit works well in the vicinity of the selected 989 time and radius, with increasing discrepancies further from the fitting point. The overall model results are shown in Fig. 4 compared with our previous results and experimental results from the Steven experiments done at Los Alamos National Laboratory and Lawrence Livermore National Laboratory (9). I Experiment - New HE | Experiment - Aged HE [J g Power Law I H Power Law II Geometry 3 1 inch HE ACKNOWLEDGEMENTS The patient and enthusiastic experimental work of Deanne Idar and Jim Straight was critical to the continued modeling. We thank Steve Chidester for sharing experimental details and results from his tests. This work is supported by the United States Department of Energy under contract W-7405ENG-36. cxpenmeiu Experiment || Power Law Calibration REFERENCES Geometry 2 1/2 inch HE 1. I Experiment Power Law I Power Law II 2. Experiment - New HE - p =1.84 Experiment - New HE - p =1.83 I Experiment - Aged HE - p =1.83 K^SSS^ Power Law I Power Law II 40 45 50 55 60 65 70 3. 75 Velocity Required For Reaction (m/s) 4. Figure 4. Comparison of models with experimental results from Steven tests. The horizontal bars show the difference between the highest no-go test and the lowest go condition for both the experiments and the models. Geometry 1 is used to calibrate the model. Geometry 2 involved a thin section of HE and there was evidence of contact between the cover plate and back support plate. The Geometry 4 experiments were done at LLNL by Steve Chidester. 5. 6. CONCLUSIONS The incorporation of the linear operator gives improved consistency in applying the ignition model, even though it does not appear to have a major effect on the results for the Steven test. We hope to expand on the model by including other dissipation mechanisms such as binder flow and void collapse. The most important need is to develop better constitutive equations for the macroscale behavior of the material. The localization behavior is key to correctly predicting the ignition behavior of real HEs. None of the currently used constitutive models correctly capture the localization effects at failure conditions. 7. 8. 9. 990 Browning, R. V., "Microstructural Model of Mechanical Initiation of Energetic Materials," in Shock Compression of Condensed Mater - 1995, edited by S. C. Schmidt et al., AIP Conference Proceedings 370, New York, 1996, pp. 405-408. McGuire, R.R. and Tarver, C.M., "Chemical Decomposition Models for the Thermal Explosion of Confined HMX, TATB, RDX, and TNT," Seventh (International) Symposium on Detonation, June 1981, pp. 56-64. Nichols, A.L. Ill, Chemical-TOPAZ - An Addendum to the TOPAZ manual, UCRL-ID-104558 Add. 1, Lawrence Livermore National Laboratory, Jan. 1993. Dienes, J.K. and Kershner, J.D., "Multiple-Shock Initiation via Statistical Crack Mechanics", Eleventh International Detonation Symposium, August 1998, pp. 717-724. Idar, D.J., et al., "Low Amplitude Insult Project: PBX 9501 High Explosive Violent Reaction Experiments," Eleventh International Detonation Symposium, August 1998, pp. 101-110. Idar, D.J. et.al.,"Low Amplitude Impact of Damaged PBX 9501," in Shock Compression of Condensed Matter - 1999, edited by M. D. Furnish et al., AIP Conference Proceedings 505, New York, 2000. Scammon, RJ. et al. "Low Amplitude Insult Project: Structural Analysis and Prediction of Low Order Reaction," Eleventh International Detonation Symposium, August 1998, pp. 111-118. Barenblatt, G. I., Scaling, self-similarity, and intermediate asymptotics, Cambridge University Press, Cambridge, 1996. Chidester, S.K., Tarver, C.M., DePiero, A.H., and Garza, R.G., "Single and Multiple Impact Ignition of New and Aged High Explosives in the Steven Impact Test", in Shock Compression in Condensed Matter - 1999, edited by M. D. Furnish et al., AIP Conference Proceedings 505, New York, 2000, pp. 663-666.
© Copyright 2026 Paperzz