采用深度学习的异步电机故障诊断方法
发布时间：2018-10-30 19:23
【摘要】：为解决传统异步电机故障诊断方法因电机结构复杂、信号非平稳和机械大数据等因素引起的诊断困难问题,提出一种高效准确的异步电机故障诊断(SDAE)方法。该方法利用堆叠降噪自编码提取信号特征,结合Softmax分类器实现高效准确的电机故障诊断。首先,采集异步电机的整体电流和振动信号,将电流信号与傅里叶变换后的振动频域信号组合构成样本,并做归一化处理;然后,构建堆叠降噪自编码网络,确定网络层数、各隐藏层节点数、学习率等参数;最后,输入训练样本依次训练自编码和分类器,微调整个网络并用测试数据验证网络的优劣。试验结果表明,在合适的参数下采用SDAE方法的异步故障诊断准确率高达99.86%,比传统电机故障诊断方法提升至少6%。
[Abstract]:In order to solve the problem of fault diagnosis of asynchronous motor caused by complex motor structure, non-stationary signal and mechanical big data, an efficient and accurate (SDAE) method for fault diagnosis of asynchronous motor is proposed. In this method, stacking noise reduction and self-coding are used to extract signal features and Softmax classifier is used to realize efficient and accurate motor fault diagnosis. Firstly, the whole current and vibration signal of asynchronous motor are collected, and the current signal is combined with the vibration frequency domain signal after Fourier transform to form a sample, and the signal is normalized. Then, the stacking noise reduction self-coding network is constructed to determine the number of network layers, the number of nodes in each hidden layer, the learning rate and other parameters. Finally, input training samples to train self-coding and classifier in turn, fine-tune the entire network and test data to verify the advantages and disadvantages of the network. The experimental results show that the accuracy of asynchronous fault diagnosis using SDAE method is as high as 99.86 under suitable parameters, which is at least 6 times higher than that of traditional motor fault diagnosis method.
【作者单位】： 南京信息工程大学信息与控制学院;南京信息工程大学计算机与软件学院;
【基金】：国家自然科学基金资助项目(51405241,51505234)
【分类号】：TM343
[Abstract]:In order to solve the problem of fault diagnosis of asynchronous motor caused by complex motor structure, non-stationary signal and mechanical big data, an efficient and accurate (SDAE) method for fault diagnosis of asynchronous motor is proposed. In this method, stacking noise reduction and self-coding are used to extract signal features and Softmax classifier is used to realize efficient and accurate motor fault diagnosis. Firstly, the whole current and vibration signal of asynchronous motor are collected, and the current signal is combined with the vibration frequency domain signal after Fourier transform to form a sample, and the signal is normalized. Then, the stacking noise reduction self-coding network is constructed to determine the number of network layers, the number of nodes in each hidden layer, the learning rate and other parameters. Finally, input training samples to train self-coding and classifier in turn, fine-tune the entire network and test data to verify the advantages and disadvantages of the network. The experimental results show that the accuracy of asynchronous fault diagnosis using SDAE method is as high as 99.86 under suitable parameters, which is at least 6 times higher than that of traditional motor fault diagnosis method.
【作者单位】： 南京信息工程大学信息与控制学院;南京信息工程大学计算机与软件学院;
【基金】：国家自然科学基金资助项目(51405241,51505234)
【分类号】：TM343
【相似文献】
相关期刊论文 前10条
1 林R，
本文编号：2300931
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2300931.html
