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

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。

Vibration and Acoustic Joint Fault Diagnosis of Conventional Circuit Breaker Based on Multi-Feature Fusion and Improved QPSO-RVM

In order to diagnose the mechanical fault of conventional circuit breaker reliably,a vibration and acoustic joint diagnosis method on the switching fault of conventional circuit breakers based on multi-feature fusion and improved quantum particle swarm optimization (QPSO)-relevance vector machine (RVM) was proposed.Firstly,the vibration signal and acoustic signal were denoised by hard and soft threshold wavelet packet denoising algorithm and the processed signals were decomposed by complementary ensemble empirical mode decomposition (CEEMD).Then the energy coefficient,sample entropy and spectrum entropy were extracted from the intrinsic mode functions to form multi feature parameters.Secondly,it reduced the dimensionality of multi feature parameters by combined kernel function kernel principal component analysis and fused the result to form the feature vector as the input of RVM,solving the problem of low recognition accuracy and low stability of single feature.Finally,the improved QPSO was used to optimize classification model parameters,and the binary tree model based on RVM was established to identify machinery fault.Experiment results show that the proposed method can effectively improve the reliability of the diagnostic results at different machineryfault condition.