Prediction of the Impact Sensitivity by Neural Networks

Abstract

A methodfor optimizing the prediction ofimpact sensitivity ofexplosive molecules hv neural networks is presented. The database we used consisted 0/272 molecules containing C,H,N,0 ofknown sensitivity and belonging to several chemicalfamilies. Pertinent molecular desriptors were selected by a preliminary multilinear treatment. The effects of the network's topology, the extent of the training, the choice of descriptors were examined and optimized. The predictions are satisfactory with a correlation coefficient of 0.94 obtained through cross-validation. Moreover 95% of compounds are correctiy classißed in a 3sensitivity scale and the remaining 5% are classißed äs ambiguous which is very encouragingfor a real worid implementation. The neural networks approach proves more accurate and more general than previous methods.

Optimization of Neural Networks Architecture for Impact Sensitivity of Energetic Molecules

“…Sets of descriptors similar to the one used by Nefati et al (10 descriptors: 1,2,5,6,7,8,14,15,16,28).…”

“…As shown in Figure 1 and 2, the increase of numbers of molecular descriptors didn't always ameliorate the results. Specifically, the NN results with S6 and S7 sets 2,2,6,9,9-Pentanitro-4,oxa-5-oxo- 6-azadecane 98 47 55 197 12_41 al25 1 ,1,1,3,6,9,11,11,11-nonanitro-3,6,9- 1,5,7,10,14,14,7, Bis-(5,5,5-trinitro-3-nitrazapentanoyl)methylenedinitramine 9 15 10 204 12_48 al27 1,1,1,4,6,6,8,11,11,1,1,3,6,6,9,11,11,9 -diazaundecane 10 10 11 206 12_50 al28 Bis-(2,2,2-trinitroethyl)-4-nitraza-1,7-heptanedioate 32 29 32 1,1,3,6,9,12,14,14,6,9, -1,1,1,3,6,6,8,10,10,13,15,15,15-tridecanitro-3,8,13- g went worse than the results with S2 and S3 sets. On the other hand, S5 set which was expanded from S1 set improved the result significantly, when compared with the one with S1 set.…”

“…Since our study was initiated by following the footsteps of Nefati et al, 8 it would be worthwhile to mention the NN architecture and results obtained by Nefati et al They employed total 204 explosive molecules among which 181 molecules were taken from the database of SSK, which is the same database we utilized. Nefati et al appeared to take a significant number of explosive molecules from other databases.…”

“…Thus, if one wants to derive a predictive methodology for impact sensitivity to cover various types of energetic molecules, neural networks (NN) may be a good choice in current circumstances. 8 …”

Shock Induced Decomposition and Sensitivity of Energetic Materials by ReaxFF Molecular Dynamics

“…Determination, evaluation and prediction of impact sensitivity have stimulated numerous studies in last decades [5][6][7][8][9][10][11][12][13]. It has been found that the impact sensitivity is influenced by various molecular parameters such as the oxygen balance [14], the molecular electro-negativity [15][16][17], the lengths of the trigger bonds [18][19][20], and the charge distribution asymmetry around these bonds both in the ground [21][22][23] and in the excited states [24][25][26][27].…”