Research In Sensitivity Prediction Of Composite Explosives

In this paper, with composition, grain size and density as affecting factors, the artificial neural network(ANN) prediction model of composite explosives sensitivity has been established. In order to get a more convenient theoretic prediction model, the sensitivity regularity has been explored and the formula has been fitted by the regression analysis method. Ultimately the prediction of composite explosives sensitive performance was realized.The predict results were as follows:(1) Most of the relative errors of the artificial neural network model of impact sensitivity, except a few at 20%, could be controlled in less than 10%.(2) To the artificial neural network of friction sensitivity, except the individual, the relative errors could be controlled at about 20%.(3) All of the relative errors of the artificial neural network model of shock sensitivity were within 5%.Respectively by linear regression analysis, the regularity of the sensitivity was studied, which was compared with the neural network model.For shock sensitivity, characterized by small scale gap test(SSGT), empirical formulas has been fitted by combining with the neural network model. Using the fitting formula to predict the shock sensitivity, the relative error was about 5 percent. According to the predicting results, it can be concluded that the artificial neural network model and the regression analysis method is feasible to predict the sensitivity of composite explosives.In this paper, the comprehensive prediction research in composite explosives was made for the first time in domestic and overseas field, which could be helpful to the research and application of composite explosives.