基于支持向量机的航空发动机振动预测模型研究

Aeroengine Vibration Forecasting based on Support Vector Machine

提出了一种基于支持向量机的航空发动机振动参数预测方法。分析了支持向量机用于时间序列预测的基本理论,对时间序列进行了相空间重构,采用互信息法计算了延迟时间,运用平均一步绝对误差选取了嵌入维数,在此基础上建立了基于支持向量机的时间序列一步预测模型。应用某发动机飞参记录数据对发动机振动参数进行预测,预测精度比RBF神经网络更高,研究结果验证了应用支持向量机模型进行发动机参数预测的正确性和可行性。

A aeroengine vibration forecasting method based on support vector machines is presented in this paper. Basic theory analysis of support vector regression in time series is introduced in detail and the reconstruction of the phase-space is presented, the delay time is obtained based on the method of mutual information and the average single-step absolute error is calculated to select embedding dimension, and then the single-step forecasting model of time series is established by using support vector machines. The support vector machines forecasting model is used to forecast aeroengine vibration by applying flight data, to compare with the result of RBF neural network, the proposed method has better forecasting precision, the research result shows the accuracy and feasibility of aeroengine vibration forecasting by applying support vector machines.