北京航空航天大学学报 ›› 2018, Vol. 44 ›› Issue (5): 923-930.doi: 10.13700/j.bh.1001-5965.2017.0353

• 论文 • 上一篇    下一篇

强噪声环境下自适应CRPF故障诊断方法

王进花1,2,3, 曹洁1,2,3, 李伟1, 黄玲1,2,3   

  1. 1. 兰州理工大学 电气工程与信息工程学院, 兰州 730050;
    2. 甘肃省工业过程先进控制重点实验室, 兰州 730050;
    3. 兰州理工大学 电气与控制工程国家级实验教学示范中心, 兰州 730050
  • 收稿日期:2017-05-24 出版日期:2018-05-20 发布日期:2018-05-29
  • 通讯作者: 曹洁 E-mail:caoj@lut.cn
  • 作者简介:王进花,女,博士研究生,副教授。主要研究方向:故障诊断、非线性滤波方法及应用;曹洁,女,教授,博士生导师。主要研究方向:智能信息处理、非线性理论及应用。
  • 基金资助:
    国家自然科学基金(61763028);甘肃省自然科学基金(1506RJZA105,1606RJZA145)

An adaptive CRPF fault diagnosis method under strong noise condition

WANG Jinhua1,2,3, CAO Jie1,2,3, LI Wei1, HUANG Ling1,2,3   

  1. 1. College of Electrical and Information Engineering, Lanzhou University of Technology, Lanzhou 730050, China;
    2. Key Laboratory of Gansu Advanced Control for Industrial Processes, Lanzhou 730050, China;
    3. National Experimental Teaching Center of Electrical and Control Engineering, Lanzhou University of Technology, Lanzhou 730050, China
  • Received:2017-05-24 Online:2018-05-20 Published:2018-05-29

摘要: 针对非线性非高斯系统在实际工作环境中受强噪声干扰影响导致的故障诊断精度低的问题,提出了一种状态转移密度方差自适应更新的代价评估粒子滤波(CRPF)故障诊断方法。通过设计观测值与先验状态之间的相关性判别函数,根据噪声和误差的大小实时自适应调整状态转移密度方差,增强算法对强噪声干扰的适应能力;研究了残差自适应阈值的设计方法,通过引入滑动窗求区间均值代替基于参数置信区间自适应阈值的均值和方差,在保证故障诊断准确性的前提下减少计算时间。以160 MW燃油机组为例,通过对不同强噪声环境下的汽包水位传感器故障诊断实例分析,结果表明该方法在复杂噪声环境下故障诊断的准确性得到了明显提高,同时减少了计算时间。

关键词: 故障诊断, 强噪声, 代价评估粒子滤波(CRPF), 自适应阈值, 漏诊率, 误诊率

Abstract: Aimed at the problem of low precision in fault diagnosis of nonlinear non-Gaussian system due to serious noise interference under the actual working condition, this paper puts forward a new fault diagnosis method, which can adaptively update the state transition density variance of a cost reference particle filter (CRPF). By designing the correlation discriminant function between the measurement value and the prior state, the variance of the state transition density was adjusted adaptively according to the magnitudes of noise and error, and the adaptability of the algorithm to strong noise interference is dramatically enhanced. Furthermore, the method for designing adaptive threshold of residual was studied, and the sliding window was also introduced to calculate the mean of interval instead of the mean and variance of the adaptive threshold based on parameter confidence interval, which was expected to reduce the calculation time under the premise of ensuring the accuracy of fault diagnosis. Taking 160 MW fuel unit as an example, drum level sensor fault diagnoses under different strong noise conditions were analyzed. From the results, it is found that the accuracy of fault diagnosis in the complex noise environment is obviously improved and the computation time is greatly reduced.

Key words: fault diagnosis, strong noise, cost reference particle filter (CRPF), adaptive threshold, missed diagnosis rate, misdiagnosis rate

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发