【作者】 刘洋；

【导师】 李凌均；

【作者基本信息】 郑州大学 ， 工程硕士（专业学位）， 2022， 硕士

【摘要】 随着现代工业的进步与发展,旋转机械的应用日趋广泛,而滚动轴承作为旋转机械的重要组成部分和易损部件,一旦发生故障将会对设备的安全稳定性起到极大的影响。因此,及时有效的进行故障诊断和模式识别研究便具有重要的现实意义。本文以滚动轴承为研究对象,使用全矢谱技术和改进优化的最大二阶循环平稳盲解卷积(CYCBD)算法相结合的方法对轴承故障数据进行预处理并提取相关故障特征,同时采用变量预测模型识别方法(VPMCD)对故障进行模式识别。主要研究内容如下:针对故障信号具有的非平稳、非线性等特点,同时为了解决最大二阶循环平稳盲解卷积(CYCBD)在滤波时需要预先设置循环频率参数的问题,本文引入了自适应局部迭代滤波(ALIF)对故障信号进行预处理,并提出了基于ALIFCYCBD的滚动轴承故障特征提取方法,通过ALIF算法对源信号进行滤波降噪预处理,采用联合系数准则选取相关分量进行重构,同时通过分析重构信号得到的估值来设置CYCBD算法分解信号时所需要的参数,最后通过CYCBD算法对重构信号进行处理分析。本文介绍了ALIF和CYCBD的算法理论和数值计算过程,并通过振动仿真信号分析和滚动轴承故障信号实验验证,同时加入对比分析,以验证所提方法在轴承故障特征提取方面的有效性。在对CYCBD算法参数的进一步分析中表明,滤波器参数的先验设置对于CYCBD的分解效果也具有一定的影响。因此,为了解决CYCBD算法能够自适应选取最优参数的问题,本文引入了基于非线性收敛改进的鲸鱼优化算法(IWOA)对滤波器长度和循环频率进行参数组合寻优,并以ALIF得到的循环频率估值设置该参数的寻优范围。传统的基于单通道的信号分析容易造成故障特征信息的遗漏。因此,本文通过引入全矢谱技术实现对同源双通道信号进行数据层面的融合,以期实现对轴承故障信号的全面描述。综上所述,本文提出了全矢谱技术和优化改进的CYCBD算法相结合的滚动轴承故障特征提取方法,在介绍了全矢谱理论以及算法流程后,通过对同源双通道轴承数据信号的实验验证,证明所提方法在抑制噪声、提高信噪比方面具有更好的效果,同时相较于单通道信号分析具有更全面的提取轴承故障信息的能力。机械故障诊断的实质是模式识别的过程,因此本文通过全矢CYCBD和基于变量预测模型分类(VPMCD)相结合的方法实现对轴承故障状态的模式识别。首先通过全矢CYCBD方法提取同源双通道信号全矢融合后的特征并组成特征向量组,并将其分为训练组和测试组,将训练组样本输入到VPM预测模型中进行训练学习,并通过测试组样本进行模式识别验证,以实现滚动轴承的模式识别过程。更多还原

【Abstract】 With the progress and development of modern industry,the application of rotating machinery is becoming more and more extensive.As an important part and vulnerable part of rotating machinery,rolling bearing will have a great impact on the safety and stability of equipment once it breaks down.Therefore,timely and effective research on fault diagnosis and pattern recognition have important practical significance.In this thesis,the rolling bearing is taken as the research object.The full vector spectrum technique and the improved and optimized Maximum Second-order Cyclostationarity Blind Deconvolution(CYCBD)algorithm are used to preprocess the bearing fault data and extract the relevant fault features.At the same time,the Variable Predictive Model Based Class Discriminate(VPMCD)is used to recognize the fault pattern.The main research contents are as follows:In view of the non-stationary and nonlinear characteristics of the fault signal,and to solve the problem that the cyclic frequency parameters need to be set in advance when the CYCBD is filtered,this thesis introduces Adaptive Local Iterative Filtering(ALIF)to preprocess the fault signal,and proposes a method for extracting the fault features of rolling bearings based on ALIF-CYCBD.The source signal is filtered and denoised by ALIF algorithm.The correlation component is selected by joint coefficient criterion for reconstruction,and the parameters needed by CYCBD algorithm for signal decomposition are set by analyzing the estimated value of the reconstructed signal.Finally,the reconstructed signal is processed and analyzed by CYCBD algorithm.In this thesis,the algorithm theory and numerical calculation process of ALIF and CYCBD are introduced,and the vibration simulation signal analysis and rolling bearing fault signal experiment verification are carried out.At the same time,comparative analysis is added to verify the effectiveness of the proposed method in extracting bearing fault features.The further analysis of CYCBD algorithm parameters shows that the prior setting of filter parameters also has a certain influence on the decomposition effect of CYCBD.Therefore,in order to solve the problem that CYCBD algorithm can adaptively select the optimal parameters,this thesis introduces the Improved Whale Optimization Algorithm(IWOA)based on nonlinear convergence to optimize the filter length and cyclic frequency,and takes the cyclic frequency estimated by ALIF as the optimization range of this parameter.Because the traditional signal analysis based on single channel is easy to cause the omission of fault feature information,this thesis introduces the Full Vector(FV)spectrum technology to realize the data level fusion of the signals of the same source and two channels,so as to realize the comprehensive description of the bearing fault signals.To sum up,this thesis puts forward a fault feature extraction method for rolling bearings,which combines the full vector spectrum technology with the optimized CYCBD algorithm.After introducing the full vector spectrum theory and the algorithm flow,the experimental verification of the same two-channel bearing data signals proves that the proposed method has a better effect in suppressing noise and improving signal-to-noise ratio,and has a more comprehensive ability to extract bearing fault information compared with single-channel signal analysis.The essence of mechanical fault diagnosis is the process of pattern recognition.Therefore,in this thesis,the pattern recognition of bearing fault state is realized by combining FV-CYCBD with VPMCD.Firstly,the FV-CYCBD method is used to extract the features of the same two-channel signals after FV fusion and form a feature vector group,which is divided into a training group and a test group.The training group samples are input into the VPM prediction model for training and learning,and the pattern recognition is verified by the testing group samples to realize the pattern recognition process of rolling bearings.更多还原