基于深度学习网络的风机传动系统主要部件故障诊断的研究

发布时间：2018-06-12 20:04

本文选题：风机传动系统 + 特征提取　；参考：《上海电机学院》2017年硕士论文

【摘要】：风机传动系统作为风机能量转换的核心部件,由于受恶劣的工作环境、风速的不确定性、交变负载等因素的影响,造成风机传动系统故障频发。本文基于深度学习网络对风机传动系统的主要部件进行故障诊断,并对风机传动系统主要部件故障诊断系统进行设计,主要研究工作如下:(1)对实验获取的轴承、齿轮振动数据进行数据预处理,利用最小二乘法消除数据的线性趋势项,再用五点三次平滑法对数据进行平滑处理,为故障诊断模型的建立奠定了基础。(2)将经过预处理但未提取特诊的轴承、齿轮数据称为原始数据,基于这些原始数据,建立基于深度学习网络的风机传动系统轴承、齿轮的故障诊断模型。对深度学习网络的关键参数进行设置,考虑不同的隐藏层数、不同样本输入长度和不同的隐藏节点数进行模型的设计,实例分析结果表明该故障诊断模型的可行性和有效性。(3)提取轴承和齿轮振动数据的特征参数,再利用深度学习网络建立风机传动系统轴承、齿轮的故障诊断模型,对轴承、齿轮的故障进行诊断。与BP神经网络和s_kohonen聚类算法进行对比试验,结果表明该故障诊断模型拥有较高的诊断准确率。(4)搭建基于深度学习的风机传动系统主要部件故障诊断的原型系统,进行系统各模块的分析。该系统包括数据预处理模型、特征提取模型、深度学习模型等,实验结果验证了所建模型的有效性。
[Abstract]:The fan transmission system, as the core component of the energy conversion of the fan, has caused the frequent failure of the fan transmission system due to the influence of the indeterminacy of the working environment, the uncertainty of wind speed and the alternating load. This paper is based on the deep learning network to diagnose the main parts of the fan transmission system and to the main part of the fan transmission system. The main research work of the fault diagnosis system is as follows: (1) preprocessing the data of the bearing and gear vibration obtained by the experiment, using the least square method to eliminate the linear trend of the data, and then smoothing the data with the five point three smoothing method, which lays the foundation for the establishment of the fault diagnosis model. (2) it will be preformed. But it does not extract the bearing of the special diagnosis. The gear data is called the original data. Based on these original data, a fault diagnosis model for the bearing of the fan transmission system based on the depth learning network is established. The key parameters of the depth learning network are set up, considering the different hidden layers, the input length of different samples and the different hidden nodes. The example analysis results show the feasibility and effectiveness of the fault diagnosis model. (3) extracting the characteristic parameters of the vibration data of the bearing and gear, and using the depth learning network to establish the bearing of the fan transmission system, the fault diagnosis model of the gear, the diagnosis of the fault of the bearing and gear, and the BP neural network and the s_kohone N clustering algorithm is used for comparison test. The results show that the fault diagnosis model has a high diagnostic accuracy. (4) build a prototype system for fault diagnosis of main parts of fan transmission system based on depth learning, and analyze the modules of the system. The system includes data preprocessing model, feature extraction model, depth learning model, and so on. The results verify the validity of the model.
【学位授予单位】：上海电机学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】