支持向量机和BP网络在Mg/PTFE性能预测中的比较  

Comparative Study on Support Vector Machine and BP Neural Network in Performance Prediction of Mg /PTFE

作　　者：赵田安[1] 欧阳的华[1] 刘振[1] 

机构地区：[1]武警工程大学装备工程学院,陕西西安710086

出　　处：《实验室研究与探索》2014年第6期60-64,共5页Research and Exploration In Laboratory

摘　　要：为了更好地预测Mg/PTFE贫氧推进剂配方与其性能之间的关系,分别采用支持向量机(SVM)和BP神经网络对Mg/PTFE贫氧推进剂的燃烧热、燃烧温度和燃速进行了预测,并将各自的预测结果与测试结果进行了比较验证。结果表明,SVM能够较好地对Mg/PTFE贫氧推进剂的性能进行预测,其预测的最大相对误差(4.2%,9.8%,10.0%)都比BP神经网络预测的相对误差(13.0%,25.9%,41.8%)小,精度较高,为Mg/PTFE贫氧推进剂的性能预测提供了一种新方法。According to the complicated relationship among the formulation design for Mg/PTFE fuel rich propellant and its combustion heat, combustion temperature and combustion rate, the support vector machine （SVM） and BP neural network in the use of performance prediction of Mg / PTFE fuel rich propellant were introduced. The results were verified by experiments at last. The results showed that the prediction maximum relative errors （4. 2% , 9. 8% , 10.0% ） of SVM were smaller than BP neural network （ 13.0% , 25.9% , 41.8% ） , and the SVM was capable of making accurate predictions of performance of the Mg/PTFE fuel rich propellant.

关 键 词：贫氧推进剂 支持向量机 BP神经网络 性能预测 

分 类 号：TJ530[兵器科学与技术—军事化学与烟火技术]