Study On Predicting Properties Of Energetic Materials

Pursuing higher energy and lower sensitivity has been the main trend for the energetic material .The research on molecule design and prediction of explosive property is becoming more and more important. To design molecule is to find the relationship between structure and property and study on it systematically through certain theory method. The program to predict the property of explosives has been established according to traditional formula and some rules in overseas. In this paper, the relationship between structure and property was studied through the quantum chemistry theory and establish the new model to predict properties of explosives by artificial neural network.59 explosive molecules have been calculated by quantum chemistry theory and obtain quantum-chemical parameters in this paper. Combined with traditional theory, the density,impact sensitivity and denotion velocity were studied respectly,and establish the predict model by artificial neural network. The precision of the model is showed as follow:(1) The relative error of prediction for density of explosives is below 5%. The value of predicting density of ONC is 2.00990g/cm3(2) The root mean squared error(RMS) of impact sensitivity is 17.84cm, and the relative RMS is 20.5,which is better than the precdicting effect of the literature [62] whose RMS is 56cm and relation RMS is 96. Compared to the two traditional model, the effect of ANN(artificial neural network)model is better.(3) The most relative error of prediction for detonation velocity is 5.35%, and the least is 0.47%. The results indicate that the model of ANN have a good precision on precdicting denotion velocity. According to the predicting results of properties of explosives,it can be concluded that the model of ANN is feasible predict the properties of explosives.It is a new way to estimate the properties of explosives,and also provide important instruct for molecule design.