摘要
风力发电机组行星齿轮箱振动信号是一种非线性非平稳的复杂信号,传统的故障诊断方法面对此类信号时,能够很好处理的范围有限。为提高在强外界干扰条件下故障智能识别的准确率,提出了一种基于一维卷积神经网络(1D-CNN)和长短期记忆网络(LSTM)混合模型的故障智能诊断方法。首先利用自参考自适应噪声消除技术(SANC)将齿轮箱振动信号分离为周期性信号分量成分和随机信号分量成分,再对包含齿轮箱故障特征的周期性信号成分进行智能特征提取和识别。经验证,所提方法较其他不同方法有明显优势,故障识别率达到99.85%,说明能有效抑制干扰信号,提高故障识别的准确率。
The vibration signal of planetary gearbox of wind turbine generator system is a kind of nonlinear and non-stationary complex signal, and the traditional fault diagnosis method cannot deal wellwith it. In order to upgrade the accuracy of intelligent fault recognition under strong external interference, anintelligent fault diagnosis method based on the hybrid model of 1 D convolutional neural network(1 D-CNN) and long-short term memory networks(LSTM).Firstly, the gearbox vibration signal was separated into periodic signal component and random signal component by using self-reference adaptive noise cancellation(SANC) technology, and then the periodic signal component including gearbox fault featureswas extracted and identified intelligently.It is verified that the method proposed in this paper has obvious advantages over other methods, and the fault identification rate reaches 99.85%, indicating that this method can effectively suppress interference signals and improve the accuracy of fault identification.
作者
黎阳羊
胡金磊
赖俊驹
王伟
赵阳
杨帆
Li Yangyang;Hu Jinlei;Lai Junju;Wang Wei;Zhao Yang;Yang Fan(Guangdong Power Grid Co.,Ltd.Qingyuan Power Supply Bureau,Qingyuan Guangdong 511500,China;College of Electrical Power Engineering,Shanghai University of Electric Power,Shanghai 200090,China)
出处
《电气自动化》
2021年第5期20-22,26,共4页
Electrical Automation
基金
广东电网有限责任公司科技项目资助(031800KK52170056)。
关键词
行星齿轮箱
自参考自适应噪声消除技术
一维卷积神经网络
长短期记忆网络
故障诊断
planetary gearbox
self-reference adaptive noise cancellation
1D convolutional neural network
long-short term memory networks
fault diagnosis
