基于多特征融合与改进QPSO-RVM的万能式断路器故障振声诊断方法

发布时间：2018-07-22 15:46

【摘要】：为可靠地进行万能式断路器机械故障诊断,在基于振动信号故障诊断的基础上,提出了一种多特征融合与改进量子粒子群(QPSO)优化的相关向量机(RVM)相结合的万能式断路器分合闸故障振声诊断方法。首先,对振声信号进行小波包软硬阈值结合去噪预处理,并利用互补总体经验模态分解算法对处理后的振声信号进行分解,提取固有模态函数能量系数、样本熵、功率谱熵,并组成多特征参数;然后,通过组合核函数核主元分析对多特征参数降维,并将其特征融合组成特征向量作为RVM的输入,解决单一特征识别断路器分合闸故障的低准确率和低稳定性;最后,利用改进QPSO优化分类模型参数,建立基于RVM的次序二叉树模型对断路器故障进行辨识。实验结果表明,该方法能有效提升不同故障状态下诊断结果的可靠性。
[Abstract]:In order to diagnose mechanical fault of universal circuit breaker reliably, based on vibration signal fault diagnosis, In this paper, a multi-feature fusion and improved quantum particle swarm optimization (QPSO) based correlation-vector machine (RVM) method for vibration diagnosis of switching faults of universal circuit breakers is proposed. Firstly, the wavelet packet soft and hard threshold is used to pre-process the vibroacoustic signal, and the complementary global empirical mode decomposition algorithm is used to decompose the vibratory acoustic signal. The energy coefficient of the natural mode function, sample entropy and power spectrum entropy are extracted. Then, by combining kernel principal component analysis, the dimension of multi-feature parameters is reduced, and the feature fusion is used as the input of RVM. In order to solve the problem of low accuracy and low stability of single feature identification circuit breaker switching fault, finally, an order binary tree model based on RVM is established to identify circuit breaker faults by using improved QPSO to optimize the classification model parameters. Experimental results show that this method can effectively improve the reliability of diagnosis results in different fault states.
【作者单位】：河北工业大学控制科学与工程学院;河北工业大学电磁场与电器可靠性省部共建重点实验室;
【基金】：河北省教育厅资助科研项目(ZD2016108)
【分类号】：TM561

【相似文献】