基于NRS-CNN的民航发动机滑油消耗量预测

Prediction of lubricating oil consumption of civil aviation engine based on NRS-CNN

针对民航发动机滑油消耗量受多个飞行阶段的多个参数影响而难以准确预测的问题,提出基于邻域粗糙集(NRS,neighborhood rough set)和卷积神经网络(CNN,convolutional neural network)的模型来预测滑油消耗量。首先,采用NRS方法提取对滑油消耗重要度较高的飞行阶段状态参数作为特征参数;然后,利用CNN对重要度高的飞行阶段状态参数进行深度特征学习,实现滑油消耗量的预测。预测结果表明:CNN能很好地完成对多滑油参数的特征提取,预测结果与实际值的平均绝对误差为0.129×10^(-3)m^(3),平均相对误差为3.8%,可满足实际工程应用的需要,为评估民航发动机滑油系统的健康状况提供参考。

With regard to the multiple parameters in multiple flight stages and the difficulty of accurately predicting lubricating oil consumption of civil aviation engines,a model based on neighborhood rough set(NRS)and convolu-tional neural network(CNN)is proposed.First,NRS method is used to extract the flight phases state parameters which are more important to the oil consumption as feature parameters;second,the CNN is used to conduct an in-depth feature study with reference to flight phase parameters so as to predict oil consumption.The results show that CNN can well complete the feature extraction of multiple oil parameters.The average absolute error between the prediction result and the actual value is 0.129×10-3 m3,and the average relative error is 3.8%,which can meet the needs of practical engineering applications and provide reference for evaluating the health status of lubricating oil system of civil aviation engine.