基于粒子滤波技术的齿轮箱故障诊断研究

发布时间：2018-06-21 14:11

本文选题：齿轮箱 + 故障诊断　；参考：《中北大学》2012年硕士论文

【摘要】：现代化生产不断朝着高性能、大规模化、高自动化方向发展。齿轮箱在机械设备中占有很重要的作用，用途比较广泛，是发生故障频率较大的部件，及时发现其故障，并对故障进行分类识别有重大的意义。但是，在齿轮箱的运行过程中存在着背景噪声，使得采集到的振动信号常常淹没在噪声中，，从而无法识别故障。为了能够准确的诊断故障，需要对采集到的振动信号进行预处理，从而提高信号的信噪比。粒子滤波技术是一种新型的基于模型的状态估计技术，在深入研究粒子滤波原理的基础上将其应用于齿轮箱振动加速度信号降噪处理中。粒子滤波方法处理降噪过程的前提是需要知道系统的状态空间模型。在本文的研究过程中，首先对采集的信号建立时序模型，文中选取的是AR模型。所需动态方程的系数确定就是利用建模过程中的模型系数。模型的阶数使用准则函数法来确定，利用最小二乘法对参数进行估计计算。本文在试验对象是JZQ250型号齿轮箱，根据实验环境及背景，以及齿轮和轴承故障故障形式及信号振动特征，利用加速度传感器采集振动信号，最后对数据进行分析处理。在上述理论分析的基础上，对实验室采集的齿轮箱振动加速度信号进行分析处理，用粒子滤波技术进行降噪，对结果比较分析发现粒子滤波降噪后的数据特征值都有所减小。神经网络是一种自适应的模式识别技术，在故障模式识别中的应用非常广泛，理论研究也较成熟。本文对经过粒子滤波降噪的数据用BP神经网络进行诊断，取得了理想效果。
[Abstract]:Modern production continues to develop in the direction of high performance, large scale and high automation. Gearbox plays an important role in mechanical equipment and is widely used. Gearbox is the component with high frequency of fault. It is of great significance to find fault in time and to classify and identify faults. However, there is background noise in the operation of the gearbox, so the vibration signals collected are often submerged in the noise, so the fault can not be identified. In order to diagnose the fault accurately, it is necessary to preprocess the collected vibration signal to improve the signal-to-noise ratio (SNR) of the signal. Particle filter is a new model-based state estimation technique, which is applied to the noise reduction of gear box vibration acceleration signal based on the in-depth study of particle filter principle. The premise of particle filter is to know the state space model of the system. In the research process of this paper, the time series model of the collected signal is established, and the AR model is selected in this paper. The coefficients of the required dynamic equations are determined by using the model coefficients in the modeling process. The order of the model is determined by the criterion function method, and the parameters are estimated by the least square method. The test object is JZQ250 gearbox. According to the experimental environment and background, gear and bearing fault forms and signal vibration characteristics, the acceleration sensor is used to collect vibration signals, and finally the data are analyzed and processed. On the basis of the above theoretical analysis, the vibration acceleration signals of the gearbox collected in the laboratory are analyzed and processed, and the noise reduction is carried out by the particle filter technique. The comparison of the results shows that the eigenvalues of the data after the reduction of the noise by the particle filter are all reduced. Neural network is an adaptive pattern recognition technology, which is widely used in fault pattern recognition, and the theoretical research is also mature. In this paper, BP neural network is used to diagnose the noise reduction data by particle filter, and the ideal results are obtained.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】