摘要
为解决非线性输出频域响应函数(NOFRF)模型用于模拟电路系统故障诊断时,传统辨识算法需多次激励计算过程耗时长的问题,提出了NOFRF的频域自适应辨识算法(NOFRF-BLMS),该算法构造了NOFRF的输入观测向量与核向量,从而可将NOFRF表示成一个伪线性结构.根据块最小均方(BLMS)原理及约束优化理论,推导出满足最小均方误差指标的NOFRF自适应辨识迭代计算公式,采用输入功率普迭代估算学习因子,由输出误差构造残差向量.NOFRF-BLMS通过在线学习方式,只需一次激励即可辨识出NOFRF,使辨识过程大幅度简化,缩短了辨识时间,具有更强的噪声抑制能力.实验结果表明,NOFRF-BLMS在相同的辨识精度下,耗时仅为传统算法的3%,且故障判断准确.
A nonlinear output frequency response functions （NOFRF） frequency adaptive identification algorithm （NOFRF-BLMS） is proposed to deal with the problems that the conventional identification method of NOFRF needs multiple stimulus and costs long time when the model of NOFRF is applied to fault diagnosis of analog circuit system. Input observation vectors and kernel vectors are constructed by means of NOFRF-BLMS, which makes the model of NOFRF become a pseudo-linear combination structure. Then, NOFRF adaptive identification recursive computational formula, which satisfies the norm of least mean square error, is deduced based on block least mean square （BLMS） and constraint optimization theory. Input power is used to estimate recursive learning factors, and output error is used to construct residual error vectors. NO-FRF is identified via online learning and only one stimulus is needed in NOFRF-BLMS which simplifies the procedure of identifying dramatically and shortens the time of identifying. NOFRF- BLMS is robust to noise. Experimental results indicate that NOFRF-BLMS costs only 3% of the identifying time of the conventional method, and the faults are correctly identified.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2011年第10期77-81,87,共6页
Journal of Xi'an Jiaotong University
基金
国家"863计划"资助项目(2008AA01Z126)
关键词
非线性输出频域响应函数
自适应辨识算法
故障诊断
nonlinear output frequency response functions
adaptive identification algorithm
fault diagnosis
