Application Research Of Genetic-neural Network Method In The Performance Prediction Of Explosive

Explosives are the major acting energy in all kinds of weapon systems,Along with thechange of the aim and the environment of war, the synthesis of new energetic materials hasbecome an important branch of weapon science. Calculation the detonation performancebefore synthesis allows us to follow the guide of the calculate data, which can greatly reduceworkload, avoid synthetic blindness, reduce cost.Therefore, explosive performance predictionplays a more and more important role in the process of energetic materials synthesis.In this paper, 11 kinds of macroscopy and microscopy factors of explosive fabricationincluding grain size, relative density, polarizability and etc. have been studied to establish theartificial neural network(ANN) prediction model of critical diameter. With the balance ofoxygen as affecting factor, aromaticity ,α-CH bond and ect. were made as command characterto establish artificial neural network model of impact sensitivity. Genetic algorithm were usedto optimize the threshold value and weight value of ANN model, in order to realize theaccurately prediction of explosive performance.This predict results were as follows:(1) the maximum error of the artificial neural network model of critical diameter was21.5%, and the minimum error was 6.84%. The GA-BP model has the maximum andminimum error of 15.5% and 0.2%, respectively.(2) the maximum error of the artificial neural network model of impact sensitivity was29.42%, while thte minimum error was 2.43%. The maximum erro of GA-BP model is10.89%, and ther minimum error was 0.According to the predicting results, ,it can be concluded that the artificial neural network model is feasible predict critical diameter and impact sensitivity of explosives. Smaller errorand better effect was obtained with the optimized GA-BP model.