摘要
航空发动机作为一种极其精密的设备,其内部传感器的运行状态决定了发动机能否稳定运行。因此,利用传感器的运行数据进行故障预测是维护发动机健康运行的关键。针对现阶段发动机故障预测精确度低的问题,提出了一种基于随机森林和时间卷积网络的混合模型。该模型利用随机森林算法进行重要性特征提取,然后添加滚动平均值和滚动标准差以增强数据特征,最后整合数据特征输入至时间卷积网络进行故障预测。采用C-MAPSS数据集进行验证,结果表明,该模型的故障预测性能相比于其他机器学习模型有较大幅度的提升。
Aero-engine is the most sophisticated equipment, and the operation state of the internal sensors determines whether it can run stably. Therefore, the use of sensor operating data for failure prediction is the key to maintaining the healthy operation of the engine. Aiming at the problem of low accuracy of engine failure prediction at the present stage, a hybrid model based on random forest and temporal convolutional network is proposed. It uses the random forest algorithm to extract important features,then adds rolling mean and rolling standard deviation to enhance data features, and finally integrates data feature into temporal convolutional network for fault prediction. The C-MAPSS data set is used for verification, and the results show that the fault prediction performance of the model is greatly improved compared with other machine learning models.
作者
王秀娜
鲁守银
任飞
Wang Xiuna;Lu Shouyin;Ren Fei(School of Information and Electrical Engineering,Shandong Jianzhu University,Jinan,Shandong 250101,China;Institute of Robotics and Intelligent Systems,Shandong Jianzhu University)
出处
《计算机时代》
2022年第10期103-107,共5页
Computer Era
基金
山东省重点研发计划(重大科技创新工程)项目(2019JZZY010435)。
关键词
航空发动机
故障预测
随机森林
时间卷积网络
aero-engine
failure prediction
random forest
temporal convolutional network