航空发动机磨损趋势RBF变权重组合预测方法研究

航空发动机滑油中金属元素的含量受许多复杂因素的影响,单一模型预测精度相对较低,本文针对这个问题提出了RBF网络变权重组合预测模型（RBFNN-VWCF）对航空发动机零部件的磨损趋势进行研究。首先引用C-C方法确定时间序列的嵌入维数和时间延时,重构相空间确定模型的输入输出样本,然后对两种模型进行组合预测,利用k均值聚类方法确定RBF网络的中心,利用固定法确定RBF网络的宽度,采用最小二乘法确定网络的权值。结果表明,RBFNN-VWCF模型充分利用了参与组合预测的两种模型的有效信息,更客观地反映发动机零部件的磨损趋势,预测结果更为稳健、相比单一模型精度更高,具有较强的工程实用价值,为发动机下一步的维修决策提供了有力支持。

RBF Variable weight combined forecasting method of aero——engine wear trend

The metal content in aero-engine spectral oil are affected by many complicated factors,wear trend prediction accuracy is much lower.To solve this problem,an RBF network variable-weight combined forecasting（RBFNN-VWCF） model is proposed in this paper to predict aircraft engine wear trend.First,phase space reconstruction by C-C method to determine the input samples and output samples of the model,and then carries on two models for combining,k-means clustering algorithm is used to determine the RBF centers,using fixed method to confirm the width of the basis function,the weights of the network is ascertained by the least square method.The results show that,RBFNN-VWCF model taking full advantage of effective information of the two models participated in the combined forecasting,can reflects the wear trend of engine much objectively.The forecasting result is robust,compares to the single model the prediction accuracy of RBFNN-VWCF model,it is tesitified higher and has strong practical value and can support powerful for next step engine decision-making.