基于双向长短时记忆网络模型的汽车搭铁失效故障诊断方法  

Fault diagnosis method based on bidirectional short and long time memory network model

作　　者：常红梅[1] CHANG Hongmei(School of Mechanical Engineering,Shaanxi University of Technology,Hanzhong 723001,China)

机构地区：[1]陕西理工大学机械工程学院,陕西汉中723001

出　　处：《兵器装备工程学报》2022年第7期286-292,共7页Journal of Ordnance Equipment Engineering

基　　金：陕西省科技厅项目(2020GY-120)。

摘　　要：现有车辆电路多数为单线制或者负极搭铁回路,为及时有效发现汽车搭铁失效故障,提出基于Bi-LSTM(双向长短时记忆网络)模型的汽车搭铁失效故障诊断方法。采用最优分数阶傅里叶变换分形维数,将汽车搭铁故障响应信号映射到分数阶空间中计算不同分数阶空间中的故障响应信号分形维数,提取故障特征向量,将其作为Bi-LSTM模型输入,通过Bi-LSTM模型标记特征向量中的异常数据细微差异,经自主学习和训练后,将所有汽车故障信息通过差异性分类,实现搭铁失效故障诊断。实验结果显示:该方法可准确得到所有故障数据变化,避免信号时频部分特性被忽略产生无法分辨现象;可以诊断更多真实异常数据,快速准确检测到故障异常点。Most of the existing vehicle circuits are single wire or negative grounding circuit,which is prone to vehicle grounding failure.In order to find the fault timely and effectively,a bi LSTM model based fault diagnosis method for vehicle grounding failure was proposed.The optimal fractional Fourier transform fractal dimension was used to map the ground fault response signal to the fractional space,calculate the fractal dimension of fault response signal in different fractional space,extract the fault feature vector,and input it as Bi LSTM model.The subtle differences of abnormal data in the feature vector were marked by Bi LSTM model.All the vehicle fault information was classified by difference to realize the fault diagnosis of vehicle grounding failure.The experimental results show that this method can accurately get all the fault data changes and avoid the phenomenon that the time-frequency characteristics of the signal are ignored;It can diagnose more real abnormal data and detect fault abnormal points quickly and accurately.

关 键 词：Bi-LSTM模型 失效故障 搭铁失效 诊断方法 分形维数 特征向量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]