基于深度自编码网络模型的风电机组齿轮箱故障检测
发布时间：2018-07-17 05:41
【摘要】：为了实现风机齿轮箱的故障检测分析,提出一种基于风电机组齿轮箱的数据采集与监视控制(SCADA)数据和振动信号的深度自编码网络模型。该模型作为一种典型的深度学习方法,通过逐层智能学习初始样本特征,可以获取数据蕴含的规则与分布特征形成更加抽象的高层表示。首先,利用限制性玻尔兹曼机对网络参数进行预训练和反向传播算法对参数进行调优,建立深度自编码网络模型。然后,通过对齿轮箱的状态变量进行编码和解码,计算重构误差并将其作为齿轮箱的状态检测量。为了有效检测重构误差的趋势变化,选用自适应阈值作为风机齿轮箱故障检测的决策准则。最后,利用对齿轮箱故障前、后记录的数据进行仿真分析,结果验证了深度自编码网络学习方法对齿轮箱故障检测的有效性。
[Abstract]:In order to realize fault detection and analysis of fan gearbox, a depth self-coding network model of data acquisition and monitoring control (SCADA) data and vibration signal based on wind turbine gearbox is proposed. As a typical depth learning method, the model can acquire the rules and distribution features of data to form a more abstract high-level representation by learning the initial sample features layer by layer intelligently. Firstly, the parameters of the network are pretrained by a restricted Boltzmann machine and the parameters are optimized by the back-propagation algorithm, and the depth self-coding network model is established. Then, by encoding and decoding the state variables of the gearbox, the reconstruction error is calculated and used as the state check measurement of the gearbox. In order to detect the trend change of reconstruction error effectively, adaptive threshold is chosen as the decision criterion for fault detection of fan gearbox. Finally, the simulation analysis of the data recorded before and after the gearbox failure is carried out, and the results show that the depth self-coding network learning method is effective for the gearbox fault detection.
【作者单位】： 华北电力大学电气与电子工程学院;中国科学院大学物理科学学院;国网冀北电力有限公司电力科学研究院;
【基金】：国家科技支撑计划项目资助(2015BAA06B03)
【分类号】：TM315
本文编号：2129258
[Abstract]:In order to realize fault detection and analysis of fan gearbox, a depth self-coding network model of data acquisition and monitoring control (SCADA) data and vibration signal based on wind turbine gearbox is proposed. As a typical depth learning method, the model can acquire the rules and distribution features of data to form a more abstract high-level representation by learning the initial sample features layer by layer intelligently. Firstly, the parameters of the network are pretrained by a restricted Boltzmann machine and the parameters are optimized by the back-propagation algorithm, and the depth self-coding network model is established. Then, by encoding and decoding the state variables of the gearbox, the reconstruction error is calculated and used as the state check measurement of the gearbox. In order to detect the trend change of reconstruction error effectively, adaptive threshold is chosen as the decision criterion for fault detection of fan gearbox. Finally, the simulation analysis of the data recorded before and after the gearbox failure is carried out, and the results show that the depth self-coding network learning method is effective for the gearbox fault detection.
【作者单位】： 华北电力大学电气与电子工程学院;中国科学院大学物理科学学院;国网冀北电力有限公司电力科学研究院;
【基金】：国家科技支撑计划项目资助(2015BAA06B03)
【分类号】：TM315
【相似文献】
相关期刊论文 前4条
1 蒙宣伊;龙辛;;风机齿轮箱故障原因浅析[J];电力技术;2010年08期
2 J.Hanna;C.Hatch;M.Kalb;卞文状;;风电机组振动监测系统对齿轮箱故障的检测与诊断[J];风能;2012年07期
3 顾煜炯;宋磊;徐天金;雷启龙;;变工况条件下的风电机组齿轮箱故障预警方法[J];中国机械工程;2014年10期
4 赵洪山;胡庆春;李志为;;基于统计过程控制的风机齿轮箱故障预测[J];电力系统保护与控制;2012年13期
相关硕士学位论文 前6条
1 张欣欣;风力发电机齿轮箱故障预警方法研究[D];燕山大学;2015年
2 高志飞;风电齿轮箱故障特征分析与诊断方法研究[D];上海电机学院;2016年
3 陈培;基于EEMD和希尔伯特平方解调技术的兆瓦级风电齿轮箱故障诊断方法研究[D];浙江理工大学;2016年
4 张伟鹏;风力发电机组齿轮箱故障特征提取方法研究[D];华中科技大学;2015年
5 马玉峰;风电机组齿轮箱故障趋势预测方法研究[D];华北电力大学;2013年
6 张燕;风电机组齿轮箱故障特征提取技术的研究[D];华北电力大学;2014年
，本文编号：2129258
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2129258.html
