中国工程机械学报

通过对航空发动机的磨损趋势进行预测，能够有效地对航空发动机磨损状态进行监测。在反映发动机磨损状态的有效观测数据中，油液分析数据能够间接反映航空发动机整体磨损趋势。因此，通过建立基于油样分析数据的趋势预测模型，从而实现发动机的磨损趋势预测。但是，目前应用于航空发动机趋势预测的模型中主要以单一预测模型为主，组合预测模型也仅是一般的线性组合，预测效果不佳。为此提出了一种基于支持向量机的非线性变权重组合预测模型，通过粒子群算法实现参数优化，油样分析数据则通过全寿命滑油系统轴承疲劳试验，间隔固定时间收集滑油样品进行性能分析得到。对其中的光谱分析数据进行组合预测分析，通过对比组合预测结果与单一预测模型的预测结果，预测精度均超过单一预测模型的预测精度，充分验证了所提组合预测模型的优越性和有效性。

By predicting the wear trend of aeroengine,the wear state of aeroengine can be monitored effectively. Among the effective observation data reflecting the engine wear state,the oil analysis data can indirectly reflect the overall wear trend of aeroengine. Therefore,by establishing a trend prediction model based on oil sample analysis data,so as to realize the wear trend prediction of engine. However,the current models used in aeroengine trend prediction are mainly single prediction models,and the combined prediction models are only general linear combinations,with poor prediction effect. Therefore,a nonlinear variable weight combination prediction model based on support vector machine is proposed,and realizes the parameter optimization through particle swarm optimization algorithm. The oil sample analysis data is obtained through the bearing fatigue test of the whole life oil system,and the oil samples are collected at fixed intervals for performance analysis. Through the combination prediction analysis of the spectral analysis data,by comparing the prediction results of the combination prediction and the prediction results of the single prediction model,the prediction accuracy exceeds the prediction accuracy of the single prediction model,which fully verifies the superiority and effectiveness of the combination prediction model proposed in this paper.