摘要
针对一类非线性动态系统,在测量数据发生传输延迟或丢失的情况下,研究了动态系统缓变故障的预测问题.重新定义了粒子滤波器的似然函数,提出了滑动窗口粒子滤波(sliding-window particle filter)算法,并得到了故障幅值的初始估计.在通过滑动窗口在线小波去噪技术对故障幅值的初始估计降噪处理后,提出了基于ARIMA模型的时间序列预测算法.上述算法能够实时地对故障幅值进行迭代估计和预测.在给定故障阈值的条件下,算法能够提前预测系统发生失效的时间.三容水箱的仿真例子说明了算法的有效性.
The likelihood function of particle filter is redefined,and sliding-window particle filter algorithm is proposed to obtain the initial estimation of the fault amplitude.After denoising to the initial estimation of the fault amplitude via on-line sliding-window wavelet denoising technology,the ARIMA model based time series prediction algorithm is brought forward.These two algorithms can iteratively estimate and predict the fault amplitude on-line.Given the threshold of failure,the two algorithms can predict the time of system failure.A simulation on 3-tank-system is given to illustrate the efficiency of the proposed techniques.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第S1期23-27,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60721003
60736026)
国家重点基础研究计划资助项目(2009CB32602)
关键词
故障预测
粒子滤波
测量丢失
传输延迟
fault prediction
particle filter
measurement missing
transmission delay
