基于蜂群优化的支持向量机齿轮故障诊断技术研究

发布时间：2018-02-22 20:41

本文关键词： 人工蜂群算法支持向量机参数优化 S变换模糊熵齿轮故障诊断　出处：《东北石油大学》2015年硕士论文　论文类型：学位论文

【摘要】：齿轮作为机械设备中的一个重要组成部分,被广泛地应用于金属切削、航天工业、国防建设、电力交通运输等各行各业中,故障发生概率比较高,一旦发生故障就会造成停机停产,甚至威胁到人员安全,因此,对其进行状态监测和故障诊断有着非常重要的实际意义。本文主要研究了基于蜂群算法优化的支持向量机齿轮故障诊断技术。本文首先研究了齿轮常见故障类型及形成原因,分析了齿轮故障振动机理,研究了常用的几种时域指标并介绍了课题使用的故障模拟试验平台QPZZ-II。其次,研究了人工蜂群算法的基本原理,针对其进化后期收敛速度较慢、种群多样性可能过分丧失、甚至陷入局部最优解的问题,引入混沌初始化和锦标赛选择策略,通过四个标准函数的寻优仿真实验验证混沌人工蜂群算法具有更快的收敛速度和更高的寻优精度;研究了支持向量机的分类方法,选择混沌人工蜂群算法对模型参数进行寻优,利用UCI数据库中的Heart、Iris和Wine数据集进行仿真实验,验证混沌人工蜂群算法优化的支持向量机具有更高的分类正确率。利用去噪前时域参数作为特征,将该方法应用于实际的齿轮故障诊断中,得到较好的故障识别效果;利用去噪后时域参数作为特征,采用CABC-SVM方法对实际齿轮进行故障诊断,对比去噪前后的识别结果,验证了时域参数的容噪能力较差。最后,研究了S变换和模糊熵的基本理论,针对齿轮信号采集过程中噪声的影响以及信号的复杂性和非平稳性,本文提出了基于广义S变换模糊熵的特征提取方法,对信号进行广义S变换得到时频对应矩阵,然后对频率分段并计算各频段的模糊熵值作为信号特征向量,通过对模拟信号的仿真实验验证该方法能够有效降低噪声的影响。将该方法应用于实际的齿轮故障诊断中,利用蜂群优化的支持向量机作为分类器,得到较高的分类正确率。
[Abstract]:As an important part of mechanical equipment, gear is widely used in metal cutting, aerospace industry, national defense construction, electric power transportation and other industries, the probability of failure is relatively high. Once a failure occurs, it will cause downtime and even threaten the safety of personnel. It is of great practical significance to monitor and diagnose its condition. This paper mainly studies the gear fault diagnosis technology of support vector machine based on bee colony algorithm. Firstly, this paper studies the common fault types and causes of gear fault. This paper analyzes the mechanism of gear fault vibration, studies several commonly used time-domain indexes, and introduces the fault simulation test platform QPZZ-III. secondly, the basic principle of artificial bee colony algorithm is studied, and the convergence speed of artificial bee colony algorithm is slow in the late stage of evolution. The population diversity may be excessively lost, or even fall into the problem of local optimal solution. Chaos initialization and tournament selection strategy are introduced. The simulation results of four standard functions show that the chaotic artificial bee colony algorithm has faster convergence speed and higher optimization accuracy, and the classification method of support vector machine is studied, and the chaotic artificial bee colony algorithm is selected to optimize the model parameters. By using Heartberg Iris and Wine data sets in UCI database, the simulation results show that the support vector machine optimized by chaotic artificial bee colony algorithm has a higher classification accuracy, and the time domain parameters before denoising are used as features. The method is applied to the actual gear fault diagnosis, and a better fault identification effect is obtained. Using the time domain parameters after denoising as the feature, the CABC-SVM method is used to diagnose the actual gear fault, and the identification results before and after denoising are compared. Finally, the basic theory of S transform and fuzzy entropy is studied, aiming at the influence of noise in the process of gear signal acquisition, the complexity and non-stationarity of the signal. In this paper, a method of feature extraction based on generalized S-transform fuzzy entropy is proposed. The time-frequency corresponding matrix is obtained by generalized S-transform. Then the frequency is segmented and the fuzzy entropy of each frequency band is calculated as the signal feature vector. The simulation results of simulated signals show that this method can effectively reduce the influence of noise. This method is applied to the actual gear fault diagnosis, and the support vector machine based on bee colony optimization is used as the classifier to obtain a higher classification accuracy.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP18;TH132.41

【相似文献】