基于振动信号复合特征向量的配电变压器机械故障检测技术研究

Distribution Transformer Mechanical Fault Detection Based on Vibration Signal Composite Feature Vector

为诊断配电变压器机械故障,提出基于振动信号复合特征向量的检测方法。针对使用单个特征量在识别故障时的信息不完整性,采集了振动信号中波形特征量和能量特征量组成复合特征向量,并使用主成分分析法将高维的复合特征向量优化得到主成分特征向量,最后利用深度置信网络对其进行配电变压器状态识别。研究结果表明:复合特征向量可以准确反映配电变压器的机械故障,使用主成分分析法的优化方法可以有效提高机器识别的工作效率,基于振动信号复合特征向量的检测方法可以实现变压器机械故障的检测。

In order to diagnose distribution transformer mechanical faults,a detection method based on vibration signal composite feature vector is proposed.In view of the information incompleteness in fault identification by using a single feature quantity,this paper collects the waveform feature quantity and energy feature quantity in vibration signal to form a composite feature vector,and uses principal component analysis method to optimize the high-dimensional composite feature vector to get the principal component feature vector.Finally,deep belief network(DBN)is used to identify the distribution transformer state.The results show that the composite feature vector can accurately reflect the mechanical fault of the distribution transformer,the optimization method of principal component analysis can effectively improve the working efficiency of the machine recognition,and the detection method based on the composite feature vector of vibration signals can realize the detection of the mechanical fault of the transformer.