基于多角度特征提取的舵机故障诊断方法

针对舵机故障中的抖动问题，提出一种基于多角度特征提取的故障诊断方法.利用短时分析法分帧舵机数据，以获得短时平稳的时间序列；引入能熵比概念提取舵机数据帧内的电流特征，并利用动态时间规整思想提取舵机数据帧内的位置特征，形成多角度特征以增强输入特征的显著性.在此基础上，利用双向长短时记忆网络提高后续舵机故障分类过程的准确性.通过某型舵机抖动实测数据的仿真，并与传统长短时记忆网络的故障诊断结果比较，验证了所提方法的有效性.

Method of Actuator Fault Diagnosis via Multiple Angles Feature Extraction

For the swaying problem of actuator faults， a fault diagnosis method based on multi-angle feature extraction is proposed. The short-time analysis idea is used to frame the actuator data to obtain a short-stable time series. The energy-entropy-ratio concept is introduced to extract the current features of actuator datum frame. Dynamic time-regular thoughts to extract positional characteristics within the actuator data frame is utilized. A multi-angle feature to enhance the significance of input characteristics is formed. On this basis， the bidirectional long-short term memory network is used to improve the accuracy of the subsequent actuator fault classification process. Finally， by the measured data from a certain type actuator swaying， the validity of the proposed method is verified by the comparing with the traditional short term memory network methods.