摘要
针对弹药自动装填系统中故障监测与排查困难的问题,以其子系统弹药协调器为例,通过4层信息抽象进行了故障诊断研究,即从真实设备到虚拟模型的抽象、从虚拟模型到响应曲线的抽象、从响应曲线到特征参数的抽象和从特征参数到故障信息的抽象。建立了协调器的不确定性模型作为样本来源,通过抽样仿真获得了样本的响应曲线族。考虑到响应曲线的连续性和平滑性,使用函数型数据分析(FDA)对响应曲线进行了特征提取。根据样本中的特征参数和不确定性参数,训练神经网络作为故障诊断机。编写了故障诊断软件,验证了诊断的可行性和有效性。
To solve the problem that fault detection and isolation is difficult in automatic ammunition loading systems,a shell transfer arm is taken as an object and the fault diagnosis is broken into four information abstractions： the abstraction from a real equipment to a simulation model,the abstraction from the simulation model to response curves,the abstraction from the response curves to feature parameters,and the abstraction from the feature parameters to fault information. The uncertainty model of the shell transfer arm is built and response curves are obtained after samplings and simulations.Considering the continuity and smoothness of signals,features are abstracted using functional data analysis（ FDA）. Neural network is trained to be a fault diagnosis machine,according to feature parameters and uncertainty parameters in samples. A fault diagnosis software is developed and its feasibility and validity is verified.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2015年第6期711-716,共6页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(51175266)
关键词
函数型数据分析
神经网络
弹药协调器
故障诊断
弹药自动装填系统
故障监测
functional data analysis
neural network
shell transfer arm
fault diagnosis
automatic ammunition loading system
fault detection
