基于ANN的丁羟复合推进剂燃速预测  

Prediction of Burning Rate of HTPB Propellant by Artificial Neural Network Model

作　　者：刘建民[1] 唐少春[2] 徐复铭[1] 周伟良[1] 

机构地区：[1]南京理工大学化工学院 [2]南京大学材料科学与工程系,江苏南京210093

出　　处：《火炸药学报》2006年第3期13-16,共4页Chinese Journal of Explosives & Propellants

基　　金：兵器预研基金资助项目(42001060203)

摘　　要：以纳米结构的F e2O3(ns-F e2O3)催化剂在RDX/AP/A l/HTPB推进剂中的燃速实验数据为基础,采用人工神经网络(ANN)中误差反传播(简称BP)算法对不同ns-F e2O3含量与推进剂燃速之间的非线性关系进行了模拟,最终确定网络结构为3-4-1型,学习速率和动量常数分别为0.75,0.45,经过56 910次迭代训练,网络收敛到均方误差为1.0×10-4。用训练好的网络对ns-F e2O3的催化作用进行了预测。结果表明,BP网络对ns-F e2O3催化作用的预测研究是可行的,除一个预测结果的相对误差较大(-9.19%)外,其他预测的相对误差绝对值均在3.5%以下,表明所建立的网络具有较好的记忆和泛化能力。Using the existing burning rate test data of RDX/AP/Al/HTPB propellants containing nano structure Fe2O3catalyst, an artificial neural network（ANN） model based on back propagation（BP） arithmetic was built to predict the burning rate of this type of propellant, and the nonlinear relation between the content of catalysts in propellants and its burning rate was simulated. As the results, the network structure was ascertained to be the type of 3-4-1, the learning rate and momentum constant were 0.75 and 0.45 respectively, and the mean square error of network arithmetic converged to 1.0×10^-4 after 56 910 iterations. The burning rate of HTPB propellant containing nano-strueture Fe2O3catalyst was predicted by the trained network model, and the output of this model had good agreement with the experiment data, there was only one with larger relative error （- 9.19% ） among prediction values, the other prediction values all had the relative error less than 3. 5%. It was suggested that the built network model has the preferable ability in aspects of remembrance and forecast to predict the burning rate of above mentioned propellants.

关 键 词：燃烧化学 人工神经网络 固体推进剂 燃烧催化剂 燃速 

分 类 号：O621.2[理学—有机化学] V512[理学—化学]