Research On The Structure Activity Relationship Of Molecular Structure And Performance Parameters

With the increasingly growing of modern national defense military technology and construction, energetic material, which is the main power source of explosives and rocket, is more and more attention by people, and its performance requirements is also more and more high. At present, acquiring energetic material depends on a large number of materials synthesis experiment over and over again. So, developing energetic material has the problems, such as high cost and low efficiency. In this paper, aiming at the problems above, using multiple linear regression, partial least squares,BP neural network algorithm to carry out the structure-activity relationship study of the polynitro aromatic compounds (PNACs),In order to composite the energetic materials of high specific performance.In this paper, specific research works are as follows:(1)According to specific density performance parameters and structure parameters, this paper study the structure-activity relations hip,then establish the density prediction model of MLR algorithm. Experiments show this method improves the accuracy of prediction of density.(2)According to specific enthalpy of formation performance parameters and structure parameters, this paper study the structure-activity relationship,then establish the enthalpy of formation prediction model of PLS algorithm.Experiments show this method improves the accuracy of prediction of enthalpy of formation.(3)According to specific detonation velocity performance parameters and structure parameters, this paper study the structure-activity relationship,then establish the detonation velocity prediction model of BPNN algorithm.Experiments show this method improves the accuracy of prediction of detonation velocity.(4)Based on the above three kinds of model, the system of molecule structure-activity analysis performance prediction was established. The system can predict density, enthalpy of formation and detonation velocity of PNACs, whose error is small and high forecast accuracy.Through checking analysis of the above three kinds of prediction model and forecasting analysis of the prediction data, we have that the error of the performance of PNACs with real and estimated values is more match, the results show that three kinds of model is more accurate; the prediction of system of molecular structure-activity analysis and performance to performance prediction effect of the PNACs is good, so the conclusion for the prediction of containing material properties provides a theoretical reference basis.