基于隐马尔科夫模型的轴承故障诊断方法研究

发布时间：2018-10-23 18:34

【摘要】：轴承是旋转类机械设备中使用的最为广泛,同时也是损坏最快的部件之一。轴承能否处于正常的工作状态直接影响着机械化生产的效率,当轴承发生故障时,机器会产生异常的振动和噪声。如何提取振动故障信号的特征值并准确地识别出故障类型是故障诊断的关键,为此本文提出了基于隐马尔科夫模型的方法建立轴承故障诊断模型。隐马尔科夫模型具有分类能力强,训练样本少,计算速度快的特点,适合于非平稳振动信号分析,通过对采集到的振动信号进行特征提取,训练具有相应状态数的隐马尔科夫模型,计算出相似概率,通过相似概率判断出对应的故障的类型。具体的研究内容如下：1.在对滚动轴承故障信号进行分析之前,首先对已提取到的轴承故障信号的特征数据进行隐马尔科夫模型的建立,并且验证了隐马尔科夫模型在轴承故障诊断中的可行性。2.提出一种基于奇异值分解与隐马尔科夫模型的故障识别方法,并建立隐马尔科夫故障识别模型。首先对滚动轴承故障信号数据进行奇异值分解,将分解得到的奇异值矩阵进行矢量量化并送入建立好的隐马尔科夫模型中进行故障类型的识别。通过实验验证奇异值分解和隐马尔科夫模型相结合的方法在轴承故障诊断中的有效性。3.提出一种基于S变换与隐马尔科夫模型的故障识别方法,并建立隐马尔科夫故障识别模型。首先对滚动轴承故障信号数据进行S变换,然后对S变换后的时频谱矩阵进行奇异值分解,最后进行矢量量化并送入建立好的隐马尔科夫模型中进行故障类型的识别。通过实验验证S变换和HMM相结合的方法在轴承故障诊断中的有效性。4.设计振动信号采集与分析系统,振动传感器采用声发射传感器PXR02,模数转换芯片采用AD9225,以STM32为主控制器以及配置相应的无线传输模块进行振动信号的采集。使用VS的串口传输功能进行数据的接收,最后通过Matlab进行振动信号数据的分析,以此来完成对轴承运行状态的实时监测与故障的诊断。
[Abstract]:Bearing is one of the most widely used and damaged parts in rotating machinery. Whether the bearing is in the normal working state directly affects the efficiency of mechanized production. When the bearing fails, the machine will produce abnormal vibration and noise. How to extract the eigenvalue of vibration fault signal and identify the fault type accurately is the key of fault diagnosis. In this paper, a method based on hidden Markov model is proposed to establish the bearing fault diagnosis model. The hidden Markov model has the characteristics of strong classification ability, few training samples and fast calculation speed. It is suitable for the analysis of non-stationary vibration signals. The hidden Markov model with corresponding number of states is trained to calculate the similarity probability and the corresponding fault type is determined by the similarity probability. The specific research contents are as follows: 1. Before analyzing the rolling bearing fault signal, the characteristic data of the bearing fault signal have been established by the Hidden Markov Model, and the feasibility of the Hidden Markov Model in the bearing fault diagnosis has been verified. 2. A fault identification method based on singular value decomposition (SVD) and Hidden Markov Model (hmm) is proposed. Firstly singular value decomposition is used to decompose the fault signal data of rolling bearing. The singular value matrix is vector quantized and sent to the established hidden Markov model for fault type identification. The effectiveness of singular value decomposition (SVD) combined with Hidden Markov Model (hmm) in bearing fault diagnosis is verified by experiments. 3. A fault identification method based on S-transform and Hidden Markov model is proposed, and a hidden Markov fault identification model is established. First, the fault signal data of rolling bearing is transformed by S transform, then the time spectrum matrix after S transform is decomposed by singular value, finally vector quantization is carried out and the fault type is identified in the established hidden Markov model. The effectiveness of S-transform combined with HMM in bearing fault diagnosis is verified by experiments. 4. 4. The vibration signal acquisition and analysis system is designed. The vibration sensor adopts the acoustic emission sensor PXR02, A / D conversion chip and uses AD9225, and STM32 as the main controller and installs the corresponding wireless transmission module to collect the vibration signal. The serial port transmission function of VS is used to receive the data. Finally, the vibration signal data is analyzed by Matlab to realize the real-time monitoring and fault diagnosis of the bearing running state.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH133.3

【相似文献】