针对多型传感器采样频率不统一,现有机器学习算法难以有效处理混频数据输入,无法充分挖掘混频信号中的设备故障特征的问题,首先提出一种混频数据输入下的长短时记 忆网络(multi-frequency long and short term memory network,MF-LSTM)架构;然后,对不同采样频率的状态数据分别进行特征提取并进行特征融合,实现混频数据输入下的电气设备的故障诊断任务;最后,利用凯斯西储大学轴承数据集对所提模型进行了算例验证,结果表明:相比于单频信号输入,混频输入平均提高故障诊断精度1.72％.该实验结果证明了所提出的基于MF-LSTM的故障诊断框架的有效性和混频数据输入的必要性.
Fault diagnosis of electrical equipment based on MF-LSTM under mixed frequency input
Due to the disparate sampling rates of various types of sensors,existing machine learning algorithms struggle to effectively process mixed-frequency data inputs,preventing the comprehensive extraction of fault features from mixed-frequency signals.To address this issue,a multi-frequency long and short term memory network(MF-LSTM)architecture was proposed firstly.And then,fea-ture extraction and merging were performed on state data at different sampling rates to achieve fault diagnosis tasks for electrical equipment with mixed-frequency data inputs.Finally,the proposed model was empirically validated using the Case Western Reserve University bearing dataset.The results show that the mixed-frequency data input improves the fault diagnosis accuracy by an average of 1.72％compared with the single-frequency data input.The experimental results prove the effective-ness of the proposed fault diagnosis framework based on MF-LSTM and the necessity of mixed fre-quency input.
electric equipmentfault diagnosisstate datafeature extractionMF-LSTM
NETL
NSTL
万方数据
