因OLTC切换时振动信号具有随机性及低频混沌特性，文献[8，9]对OL

Due to the randomness and low-frequency chaotic characteristics of the vibration signal during OLTC switching, the literature [8, 9] reconstructed the phase space of the OLTC vibration signal, identified the operating conditions of the OLTC by defining the phase space coefficients, and proposed the power OLTC fault diagnosis model based on spectrum and invisible Marco chain. However, due to the limitations of the phase space reconstruction algorithm itself, the chaotic information of the signal cannot be fully presented. Literature [10] introduced the narrow-band noise-assisted multivariate EMD algorithm into the mechanical fault monitoring of the converter OTLC. It mainly suppressed the modal aliasing phenomenon in the EMD algorithm by adding narrow-band noise signals to overcome some of the defects of the traditional methods, and Higher diagnostic efficiency. However, the method for determining the feature matrix used in this method is slightly simpler and needs further study

因OLTC切换时振动信号具有随机性及低频混沌特性，文献[8，9]对OLTC振动信号进行相空间重构，通过定义相空间系数对OLTC的工况进行识别，并提出了基于振动信号功率谱与隐形马尔科链的OLTC故障诊断模型。然而由于相空间重构算法本身的局限性，使得信号混沌信息无法全部呈现。文献[10]将窄带噪声辅助多元EMD算法引入换流变OTLC的机械故障监测中，其主要是通过加入窄带噪声信号来抑制EMD 算法中的模态混叠现象克服了传统方法一些缺陷，且具有较高的诊断效率。但该方法所用的特征矩阵的判断方法略为简单，需要进一步研究

Due to the randomness and low-frequency chaos of the vibration signal during OLTC switching, the paper [8,9] reconstructs the phase space of the OLTC vibration signal, identifies the working condition of OLTC by defining the phase space coefficient, and proposes the OLTC fault diagnosis model based on the power spectrum of the vibration signal and the hidden Markov chain. However, due to the limitations of the phase space reconstruction algorithm, the chaotic information of the signal can not be fully presented. In reference [10], the narrow-band noise assisted multivariate EMD algorithm is introduced into the mechanical fault monitoring of the converter OTLC, which mainly uses the narrow-band noise signal to suppress the modal aliasing phenomenon in the EMD algorithm, overcomes some defects of the traditional method, and has high diagnosis efficiency. However, the judgment method of the characteristic matrix used in this method is slightly simple and needs further study<br>