摘要
准确检测飞机即将发生的故障或预测其状态的变化趋势,对于实现飞行安全具有重要意义。文章针对传统基于回声状态网络在故障预测中的不足,构建了基于小波降噪的回声状态网络预测模型。该模型保留了非线性时间序列回声状态网络预测的优势,并采取小波变换对混沌时间序列进行降噪预处理,有效提高了含噪混沌时间序列的预测精度。论文通过对某飞机发动机滑油散热器温度时间序列数据序列进行预测分析,表明文中模型具有较好的预测精度,验证了模型的有效性。
It is significant for flight safety to accurately detect the coming fault of aircraft or predict its change trend. Aiming at suppressing the shortcoming of fault prediction based on traditional ESN, we present a new predic- tion method combining ESN with wavelet denoising. Sections 1 and 2 of the full paper explain our prediction model mentioned in the title, which we believe is effective and whose core is： the method not only reserves the advantages of ESN model in nonlinear time series prediction but also reduces the noise influence in practice, i. e. , the pre- treatment via wavelet transform will be done before prediction. Section 3 concerns a certain type of aero-engine lu- bricator. Its simulation results are presented in Figs. 4,5,7,8 and Tables 1 and 2. The simulation results and their analysis show preliminarily that the proposed method improves the prediction accuracy of nonlinear chaotic time se- ries including noises, thus indicating that the proposed model is an effective approach in actual application.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2012年第4期607-611,共5页
Journal of Northwestern Polytechnical University
基金
国家重点基础研究发展计划(973计划)
国家自然科学基金(61001023
61101004)
航空科学基金(2010ZD53039)
陕西省自然科学基金(2010JQ8005)资助
关键词
小波变换
回声状态网络
非线性混沌时间序列
故障预测
aircraft, efficiency, errors, mathematical models, noise abatement, wavelet transforms
echo state networks （ESN） , nonlinear time series, prediction
