【作者】 孙磊；

【导师】 郭正刚；

【作者基本信息】 大连理工大学 ， 机械设计及理论， 2023， 硕士

【摘要】 滚动轴承是机械设备重要的旋转零件,也是机械设备的主要故障源之一。滚动轴承状态监测及故障诊断是轴承运行及维护的常用手段,对保证轴承的安全稳定运行具有重要意义。但是,在滚动轴承的状态监测过程中,由于工况的复杂性,监测信号中的噪声干扰及设备的负载变化,均对诊断的准确率有较大的影响。为了提高复杂工况下的轴承故障诊断的准确率,本文基于深度学习理论,以滚动轴承为研究对象,以噪声干扰、变负载工况下的轴承故障诊断为应用场景,提出两种轴承智能故障诊断算法。具体研究内容如下:(1)针对噪声干扰下滚动轴承故障特征提取困难的问题,提出了FSWT-DRSN算法。首先将一维振动信号通过FSWT变换为二维时频图,作为算法的输入。接着引入残差连接、注意力机制和软阈值对卷积神经网络进行改进,搭建深度残差收缩网络(DRS N)。残差连接有效解决网络退化问题,注意力机制和软阈值能根据输入数据对诊断结果贡献大小自适应设置噪声阈值,降低了噪声干扰对故障诊断的影响。最后在不同强度噪声干扰下进行对比试验。结果表明,本文提出的FSWT-DRSN算法平均诊断准确率比CNN、Res Net分别提高了8.34%、6.46%,算法具有较强的抗噪性能。(2)针对变负载工况使数据分布产生显著差异的问题,提出多尺度卷积子域适应算法。构建不同负载下的数据集作为源域和目标域,利用不同尺度大小的卷积核并行计算,充分提取源域故障特征的局部和全局信息。将局部最大均值差异作为优化目标,缩小了不同负载下数据的分布差异,实现跨负载工况下的轴承故障诊断。最后设计六组变负载工况迁移试验。结果表明,提出算法在不同负载下均能取得93.5%以上准确率,验证算法具有较强的适应性。(3)开发滚动轴承故障诊断系统并设计噪声干扰、变负载等复杂工况下的滚动轴承故障诊断试验。验证了在噪声干扰和变负载条件下,FSWT-DRSN算法及多尺度卷积子域适应算法的有效性。上述研究为噪声干扰、变负载等复杂工况下的轴承智能故障诊断提供了解决方案,并开发轴承故障诊断系统,为算法提供工业应用平台,具有一定的理论和应用价值。更多还原

【Abstract】 Rolling bearing is an important rotating part of mechanical equipment,and also one of the main failure sources of mechanical equipment.Rolling bearing condition monitoring and fault diagnosis are common means of bearing operation and maintenance,which is of great significance to ensure the safe and stable operation of bearings.However,in the process of rolling bearing condition monitoring,due to the complexity of working conditions,noise interference in monitoring signals and load changes of equipment,all have a great impact on the accuracy of diagnosis.In order to improve the accuracy of bearing fault diagnosis under complex working conditions,based on deep learning theory,this thesis takes rolling bearings as the research object and bearing fault diagnosis under noise interference and variable load conditions as the application scenario,and proposes two kinds of bearing intelligent fault diagnosis algorithms.Specific research contents are as follows:(1)FSWT-DRSN algorithm was proposed to solve the difficulty of fault feature extraction of rolling bearings under noise interference.Firstly,one-dimensional vibration signal is transformed into two-dimensional time-frequency graph by FSWT,which is used as the input of the algorithm.Then,the residual connection,attention mechanism and soft threshold are introduced to improve the convolutional neural network,and the deep residual shrinkage network(DRSN)is built.The residual connection can effectively solve the problem of network degradation,and the attention mechanism and soft threshold can adaptively set the noise threshold according to the input data contribution to the diagnosis result,thus reducing the influence of noise interference on fault diagnosis.Finally,a comparative test was carried out under different intensity noise interference.The results show that the average diagnostic accuracy of the proposed FSWT-DRSN algorithm is 8.34% higher than that of CNN and6.46% higher than that of Res Net,respectively,and the algorithm has strong anti-noise performance.(2)A multi-scale convolution subdomain adaptive algorithm is proposed to solve the problem of significant differences in data distribution under variable load conditions.Data sets under different loads were constructed as source domain and target domain,and convolution kernels of different scales were used for parallel computation to fully extract local and global information of fault features in source domain.Taking the local maximum mean difference as the optimization objective,the distribution difference of data under different loads was reduced,and bearing fault diagnosis under cross-load conditions was realized.Finally,six groups of migration tests under variable load conditions were designed.The results show that the proposed algorithm can achieve more than 93.5% accuracy under different loads,which verifies that the algorithm has strong adaptability.(3)Develop rolling bearing fault diagnosis system and design rolling bearing fault diagnosis tests under complex working conditions such as noise interference and variable load.The effectiveness of FSWT-DRSN algorithm and multi-scale convolution subdomain adaptive algorithm is verified by experiments under the condition of noise interference and variable load.The above research provides a solution for intelligent bearing diagnosis under complex conditions such as noise interference and variable load,and develops a bearing fault diagnosis system to provide an industrial application platform for the algorithm,which has certain theoretical and application value更多还原