【作者】 张澎涛；

【导师】 刘晋浩；

【作者基本信息】 东北林业大学 ， 机械设计及理论， 2014， 博士

【摘要】 齿轮箱作为机械设备中一种必不可少的连接和传递动力的通用零部件,在金属切削机床、航空、电力系统、农业机械、运输机械、冶金机械等现代工业设备中得到了广泛的应用。作为传动机械,其运行状态好坏与否,直接影响到整个机械传动组的工作性能。因此研究齿轮箱故障诊断技术和方法,对齿轮箱进行状态检测及故障分析诊断、确保齿轮箱正常工作具有重要意义。齿轮箱故障诊断技术是通过分析处理齿轮箱运行时的状态信息。定量识别其技术状态,并诊断异常故障状态的一门多学科交叉的综合技术。齿轮箱振动信号中包含了大量的工作状态信息,利用齿轮箱振动信号诊断故障是一种行之有效的方法。应用振动信号分析法对齿轮箱进行故障诊断的过程中,滤波去噪处理、故障特征提取是其中尤为重要的两个问题,一直被公认为是故障诊断中的关键环节。本文从工程实际应用的角度出发,分别以风力发电机行星齿轮箱和通用工业齿轮箱故障振动信号为具体研究对象,综合应用谱峭度、黄金分割、下山单纯形、小波分析、快速傅里叶变换、粒子群优化、冲击响应谱、瞬态分析等理论,基于混合优化理念,对齿轮箱故障振动信号去噪及故障特征提取算法两个关键问题进行了系统地研究,为齿轮箱故障诊断技术的开发研究提供了一定的理论支持。本文重在研究齿轮箱故障振动信号的优化滤波去噪和故障脉冲瞬态特征提取算法。(1)分析了齿轮箱振动信号常用分析方法的基本原理及适用范围,为后续的振动信号去噪、故障特征提取算法研究提供一定的理论支持。(2)为了解决单一的黄金分割法优化速度较慢的问题,将黄金分割和抛物线插值两种算法相结合构成加速一维搜索算法。(3)为了解决单一的传统多维优化算法收敛速度较慢,且收敛时容易陷入局部极值的问题。以峭度最大值为目标函数,根据振动信号频域分析和小波分析的原理,基于一维搜索-多维搜索(参数粗调-参数微调)混合优化的自适应滤波算法,分别采用了两种不同的滤波手段。切比雪夫带通滤波和Morlet小波滤波。应用混合优化算法,即谱峭度-加速一维搜索算法-下山单纯形法分别优化切比雪夫带通滤波器和Morlet小波滤波器的设计参数,对齿轮箱故障振动信号进行滤波去噪仿真处理。(4)为了对比不同类型混合优化算法的优劣性,将以下四种优化算法：谱峭度、黄金分割、下山单纯形、遗传算法,按一维搜索-多维搜索模式混合优化切比雪夫带通滤波器参数,并进行去噪仿真实验。(5)在前期研究的基础上将粒子群优化算法应用到齿轮箱故障振动信号去噪中,将基于加速一维搜索-粒子群优化的混合优化算法用于优化切比雪夫带通滤波器和Morlet小波滤波器的相关参数,并对齿轮箱故障振动信号进行滤波去噪仿真处理。(6)为了提取能够反映齿轮箱工作信息变化情况以及故障未来发展趋势的故障脉冲瞬态特征,将冲击响应谱分析和瞬态分析法应用到齿轮箱齿轮的故障特征提取中,用于提取衡量齿轮箱故障严重程度的三个瞬态特征指标：冲击响应谱指标SRS以及齿轮啮合系统的固有频率ωn、振荡阻尼比ζ。更多还原

【Abstract】 As an essential common component which can interconnect and transmit the power in mechanical equipment,gearbox has been widely used in modern industrial equipment,such as metal cutting machine tool, aviational and powerful system, agricultural machinery, transport machinery,metallurgical machinery. As mechanical drive,the running state of gearbox directly affect the performance of mechanical drive system. So that it is very important to research the technology and method about fault diagnosis of gearbox,detect state and diagnose the fault of gearbox,and ensure gearbox in normal working state.Fault diagnosis of gearbox is state information by analyzing and processing the gearbox is operated, the quantitative recognition technology state, and amulti-disciplinary comprehensive technology in diagnosis of abnormal fault state. Vibration signals of the gearbox contains the working state of a huge amount of information, the fault diagnosis of gearbox vibration signal is a kind of effective method. Application of vibration signal analysis and fault diagnosis of the gearbox method, denoising, feature extraction and recognition of fault type are three of the most important problems, especially in fault feature extraction, it directly affects the accuracy of diagnosis results, has been recognized as the key link in fault diagnosis. From the point of engineering application, this thesis systematically investigated the fault feature extraction and fault vibrational signal denoising methods of the diesel engine applying spectral kurtosis, golden section, Nelder-Mead, wavelet analysis,FFT,PSO,SRS,SOT,etc.This thesis mainly studies the optimal filtering vibration signal of gearbox fault denoising and fault transient pulse feature extraction algorithm.The main contents are as follows:(1) This thesis analyses the principle of the vibrational signal analysis method of gearbox to provide theoretical support for the follow-up study on vibration signal denoising and fault feature extraction algorithm.(2) In order to solve the problem that the single golden section has slow speed this thesis presents an accelerated one-dimensional search algorithm.(3) In order to solve the problem that the single multi-dimensional algorithm has slow convergence speed and easily fall into local extremum,where the objective function is kurtosis maximum, this thesis presents an adaptive filtering algorithm based on one-dimensional search-multi-dimensional search method (parameter coarse-parameter tuning). According to the principle of frequency domain analysis of vibration signal analysis and wavelet transform,, using two different filtering methods, Chebyshev bandpass filtering and Morlet wavelet filtering.Parameters of Chebyshevbandpass filter and Morlet wavelet filter are separately optimized by spectral kurtosis-golden section-Nelder-Mead and golden section-Nelder-Mead,then denoising the fault vibrational signal of gearbox.(4) In the simulation study process of one dimension-multidimensional search hybrid optimized algorithm,according to the combination of one dimensional searchand multi-dimensional search we use several different algorithms to optimizing the parameters,such as spectral kurtosis,GA, golden section and Nelder-Mead.(5) On the basis of previous study the particle swarm optimization algorithm is applied to denoising the fault vibrational signal of gearbox.This thesis proposed a hybrid algorithm that effectively reduce the searching range of PSO, and save the optimized time based on accelerated one dimensional search and PSO.lt is used for parameters optimization of Chebyshev band-pass filter and Morlet wavelet filter,then denoising the fault vibrational signal of gearbox.(6) In order to transient fault feature extraction that can reflect gearbox information changes and future development trend of the fault,the shock response spectrum analysis and the transient analysis are applied to extract fault feature of gear in the gearbox, and successfully extracting three transient characteristics indexes which can measure the severity of fault, respectively shock response spectrum index SRS,natural frequency con and Oscillation damping ratio ζ in the gear meshing system.更多还原