齿轮箱故障诊断中粒子滤波技术的应用

发布时间：2018-10-10 14:21

【摘要】：齿轮箱作为机械设备中的一个重要组成部分,对其进行状态监测和故障诊断有着重要的意义。由于齿轮箱结构复杂,工作环境比较恶劣,当发生故障时,其振动信号通常表现出非平稳特性,并且故障特征信息往往湮没在强大的背景噪声中,因此,在对其进行故障分析前需要进行降噪处理。粒子滤波是一种基于蒙特卡罗模拟和递推贝叶斯估计的滤波方法,它通过模拟方程由量测空间递推到状态空间,采用粒子描述状态空间,用由粒子及其权重组成的离散随机测度近似真实的状态后验分布,并且根据算法递推更新离散随机测度。它可以处理模型方程为非线性、噪声分布为非高斯分布的问题。本文在深入研究粒子滤波理论的基础上,利用粒子滤波技术对齿轮箱振动加速度信号进行降噪处理。建立了齿轮箱振动信号的ARMA模型,进而转化成对应的状态空间模型,把降噪问题转化成在状态空间模型下的滤波问题。在上述理论分析的基础上,本文以齿轮箱为实验对象,通过在选好的测点上安装加速度传感器来采集齿轮箱振动信号,然后对采集到的信号运用粒子滤波技术进行降噪,并通过仿真和实测信号对其进行了验证,最后,针对非线性系统的故障检测问题,提出了一种基于粒子滤波状态估计的非线性系统故障检测算法,并将该算法应用到齿轮箱上,取得了一定的成果。
[Abstract]:As an important part of mechanical equipment, gearbox is of great significance for condition monitoring and fault diagnosis. Because the structure of gearbox is complex and the working environment is very bad, the vibration signal of gearbox usually shows non-stationary characteristic when it breaks down, and the fault characteristic information is often buried in the strong background noise. Noise reduction is needed before fault analysis. Particle filter is a filtering method based on Monte Carlo simulation and recursive Bayesian estimation. The discrete random measures composed of particles and their weights are used to approximate the real state posteriori distribution, and the discrete random measures are updated recursively according to the algorithm. It can deal with the problem that the model equation is nonlinear and the noise distribution is non-Gao Si distribution. Based on the research of particle filter theory, the noise reduction of gear box vibration acceleration signal is carried out by using particle filter technology in this paper. The ARMA model of the gear box vibration signal is established, and the corresponding state space model is transformed into the corresponding state space model. The problem of noise reduction is transformed into the filtering problem under the state space model. On the basis of the above theoretical analysis, this paper takes the gearbox as the experimental object, collects the vibration signal of the gearbox by installing the acceleration sensor on the selected measuring point, and then uses the particle filter technology to reduce the noise of the collected signal. Finally, a fault detection algorithm for nonlinear system based on particle filter state estimation is proposed, and the algorithm is applied to the gearbox. Some achievements have been made.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【相似文献】