一种战车主减速器温度预测方法研究  

Research on Temperature Prediction Method of Main Reducer for Chariot

作　　者：李田科[1] 沙卫晓[1] 李伟[1] 于仕财[2] 

机构地区：[1]中国人民解放军91980部队 [2]海军航空工程学院

出　　处：《装备环境工程》2016年第6期35-40,共6页Equipment Environmental Engineering

基　　金：国家自然科学基金(61179017)~~

摘　　要：目的针对战车主减速器温度预测需求,建立时间序列ARIMA多步预测和BP神经网络预测模型,提出基于BP神经网络修正误差的ARIMA模型温度预测方法。方法结合BP神经网络的非线性能力与ARIMA模型预测能力,分析ARIMA在多步预测时误差产生原因,在神经网络对ARIMA多步误差进行预测基础上计算修正因子,把误差修正因子和BP网络结合,实现对多步预测误差的修正。结果ARIMA模型多步预测时,预测误差随预测步数的逐步增加不断增大,引入了误差修正因子进行修正。通过预测值与实际值进行对比,可有效提高预测准确度。结论 BP神经网络和误差修正因子结合应用可显著提高温度预测效果。Objective To establish models for time series ARIMA multistep prediction and BP neural network prediction as required by temperature prediction of main reducer for chariot and propose methods for temperature prediction of ARIMA model based on BP neural network error correcting. Methods Nonlinear ability of BP neural network and prediction ability of ARIMA model were combined to analyze causes of error in multistep prediction of ARIMA. Correction factors were calculated based on ARIMA multistep error prediction of neural network. Error correction factors and BP neural network were combined to achieve correction of error correction factors. Results The prediction error increased with prediction steps in multistep prediction of ARIMA model. Error correction factors were used for correction. Comparison of prediction values and actual values improved the prediction accuracy effectively. Conclusion Combined application of BP neural network and error correction factors can improve the temperature prediction result effectively.

关 键 词：误差修正因子 温度预测 ARIMA模型 BP神经网络 

分 类 号：TJ812.6[兵器科学与技术—武器系统与运用工程]