摘要
为避免关键部件故障带来的铁路货车运行安全隐患,提出基于深度学习算法的铁路货车关键部件故障检测方法。使用货车故障动态图像检测系统采集铁路货车关键部件故障图像,采用主成分分析算法提取故障图像的特征向量,利用基于深度置信网络的关键部件故障检测方法,深入分析网络基本原理和结构,将提取的特征向量作为网络的输入数据,通过网络预训练和参数更新实现铁路货车关键部件故障检测。实验结果表明:该方法能精准获取铁路货车关键部件故障图像,且较为清晰;提取的各类型关键部件故障特征均未出现重叠现象,可分性良好;不同类型关键部件故障误检率和漏检率分别低于9%、5%;不同噪声水平下,各类型关键部件故障检测平均绝对误差始终在0.3以下。
In order to avoid the hidden danger of railway freight train operation safety caused by key component faults,a fault detection method of railway freight train key components based on deep learning algorithm is proposed.The fault images of key components of railway freight cars are collected by the freight car fault dynamic image detection system,the feature vector of the fault image is extracted by the principal component analysis algorithm,the basic principle and structure of the network are deeply analyzed by using the key component fault detection method based on the deep confidence network,and the extracted feature vector is used as the input data of the network.The fault detection of key components of railway freight cars is realized through network pre training and parameter updating.The experimental results show that this method can accurately obtain the fault image of key components of railway freight cars,and it is clear;The extracted fault features of various types of key components do not overlap and have good separability;The false detection rate and missed detection rate of different types of key components are lower than 9%and 5%respectively;Under different noise levels,the average absolute error of fault detection of various types of key components is always below 0.3.
作者
丁凤霞
刘振中
DING Feng-xia;LIU Zhen-zhong(CHN Energy Investment Group Co.,Ltd.,Beijing 100120 China;Beijing Jingtianwei Technology Development Co.,Ltd.,Beijing 100085 China)
出处
《自动化技术与应用》
2023年第10期38-41,152,共5页
Techniques of Automation and Applications
关键词
深度学习算法
铁路货车
关键部件
故障检测
主成分分析
deep learning algorithm
railway wagons
key components
fault detection
principal component analysis
