基于支持向量机方法的发动机性能趋势预测

Study of support vector machines for aeroengine performance trend forecasting

关键词:  航空发动机  性能预测  时间序列预测+  支持向量回归+  模型

Key words:  Aircraft engine  Performance prediction  Time series forecasting~+  Support vector machines~+  Model

摘要:

为了提高对航空发动机性能趋势预测的精度,提出利用支持向量机方法来预测表征发动机整体性能的参数—性能综合指数。建立了基于支持向量回归的一步及多步预测模型,利用该模型对性能正常衰退及性能异常发动机的综合指数分别进行预测,并与自回归(AR)模型的预测值进行比较。结果表明,基于支持向量机的预测模型比AR模型的预测精度更高,其四步预测精度由80·56%提高到88·51%。因此该模型尤其适合中、长期预测。

Abstract:

In order to improve forecasting accuracy of aeroengine performance, a novel forecasting method based on support vector machines was presented and used to forecast performance synthetic exponent characterized the whole performance of engine. Therefore, a single-step and a multi-step forecasting model based on support vector machines was built and used to forecast performance synthetic exponent of a normal and an abnormal engine. Moreover, forecasting results are compared to those of AR model. The comparison results show that accuracy of forecasting model based on support vector regression is higher than that of AR model. For example, its four step forecasting accuracy is increased from 80.56% to 88.51%. It is suggested that forecasting model based on support vector regression is especially suitable to middle intervals or long intervals forecasting of aeroengine performance.