摘要
由于现有诊断方法的振动波形幅度普遍较大,呈现多种异常诊断结果,为此研究基于多源数据融合与机器学习算法的电气柜故障诊断。针对数据低能耗采集的需求,根据传感器节点的部署高效收集信息。在提取电气设备的故障信号特征后,对电气柜的故障数据进行特征融合。利用机器学习算法挖掘故障类别,获得节点运行数据之间的关联性,再通过Softmax分类器输出故障诊断结果。实验结果表明,实验组在数据量达到600个样本点时,振动波形振幅突然异常降低,降至-2 mm以下。这一显著变化明确指示了该点存在故障,体现了诊断过程的准确性和有效性。
Due to the generally large amplitude of vibration waveforms and the presentation of various abnormal diagnostic results in existing diagnostic methods,this study investigates the fault diagnosis of electrical cabinets using multi-source data fusion and machine learning algorithms.To meet the demand for low-energy data collection,efficiently collect information based on the deployment of sensor nodes.After extracting the fault signal features of electrical equipment,perform feature fusion on the fault data of electrical cabinets.Using machine learning algorithms to mine fault categories and obtain correlations between node operation data,the model outputs fault diagnosis results through a Softmax classifier.The experimental results showed that when the data volume reached 600 sample points,the amplitude of the vibration waveform suddenly decreased abnormally,dropping below-2 mm.This significant change clearly indicates the existence of a fault at that point,reflecting the accuracy and effectiveness of the diagnostic process.
作者
陆飞
LU Fei(East China Air Traffic Management Bureau of Civil Aviation of China,Shanghai 200335,China)
出处
《电工技术》
2025年第3期164-166,共3页
Electric Engineering
关键词
多源数据融合
机器学习
电气柜
故障诊断
multi-source data fusion
machine learning
electrical cabinet
fault diagnosis
