摘要
提出了一种针对挖掘机液压系统的非线性有源自回归(nonlinear auto-regressive with extrainputs,NARX)网络模型的故障检测方法。NARX网络模型是一种将有源自回归(auto-regressivewith extra inputs,ARX)模型与神经网络相结合的系统建模方法,具有很强的非线性辨识能力。该方法首先选取合理的网络模型结构,并根据AIC准则确定最佳模型阶数;使用正常状态样本对NARX网络进行训练,建立系统的辨识模型;然后运用序贯概率比检验(sequential probability ratiotest,SPRT)对NARX辨识模型的残差进行假设检验,检测系统的故障状态。实验分析表明,基于NARX网络模型的故障检测方法能够有效地应用于挖掘机液压系统。
A fault detection approach of excavator hydraulic system based on nonlinear auto-regressive with extra inputs （NARX） network model is proposed. As a combination of auto-regressive with extra inputs （ARX） and neural networks, NARX network is suitable for identifying nonlinear systems. Firstly, an appropriate network structure is chosen and the optimal orders of the model are obtained according to Akaike＇ s information criterion （AIC）. Then, a NARX network model is trained to develop an identification model. Finally, sequential probability ratio test （SPRT） is performed for a hypothesis test of the model residual for detecting the system faults. Experimental results show that the proposed approach based on NARX network model is effective for fault detection of an excavator hydraulic system.
出处
《机械科学与技术》
CSCD
北大核心
2008年第7期937-940,944,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
国家"863"高技术研究发展计划项目(2003AA430200)资助
关键词
液压系统
挖掘机
故障诊断
非线性有源回归网络模型
序贯概率比检验
hydraulic system
hydraulic excavator
fauh diagnosis
nonlinear auto-regressive with extra inputs
sequential probability ratio test
