航空发动机油样光谱分析的PSO-LSSVM组合预测方法

油样光谱分析是航空发动机磨损状态监测与故障诊断的重要技术,基于光谱数据的航空发动机状态预测有利于发现航空发动机的早期磨损故障。根据光谱数据特征,选取AR模型、BP神经网络模型以及GM(1,1)预测模型作为基础模型,建立了基于最小二乘支持向量机的组合预测模型,同时,用粒子群算法对LSSVM的正则化参数以及核函数参数进行了优化。最后利用两组实际的航空发动机光谱分析数据对模型进行了验证,与基础模型的对比结果充分表明,提出的带粒子群优化的最小二乘支持向量机(the Least Squares Support Vector Machines with Particle SwarmOptimization-PSO-LSSVM)的非线性变权重组合预测模型具有更好的预测精度。

Combinational Forecast Method Based on PSO-LSSVM in Spectrometric Oil Analysis of the Aircraft Engine

The spectrometric oil analysis(SOA) is an important technique for aircraft engine state monitoring and fault diagnosis,and forecasting aircraft engine state through SOA results has an advantage of finding out aircraft engine wear fault early.According to the characteristics of the SOA data,the combinational forecast model was set up based on the least squares support vector machine after Auto Regressive(AR) model,GM(1,1) model and back propagation(BP) neural network model.In addition,the particle swarm algorithm was used to optimize the regularization parameter of least squares support vector machines(LSSVM) and the parameter of kernel function.Finally,two time series of SOA data were used to verify this model.By comparying with the foundation models,the result of combinational forecasting model shows better effect and higher precision of forecast by using the non-linear variable weight and the least squares support vector machines with particle swarm optimization.