面向风电机组的齿轮箱轴承故障诊断技术研究

发布时间：2018-11-16 08:09

【摘要】：齿轮箱是风电机组传动系统的重要组成部分,齿轮箱的滚动轴承是传动链中故障率较高的部件之一。轴承发生故障不但影响机组的正常运行,甚至波及供电侧电网的安全平稳运行。因此对风电机组齿轮箱轴承故障进行快速诊断具有重要的现实意义和使用价值。本文在深入分析风力发电机组的基本组成结构、故障机理、故障特征及其特征频率的基础上采用经验模态分解(Empirical Mode Decomposition)改进阈值方法对滚动轴承出现故障时的振动信号进行降噪预处理,然后提取故障特征,分析故障类型。将时频域分析技术和旋转机械故障理论知识结合起来,针对风力发电机组的机械传动系统出现的常见机械故障问题,如断齿、点蚀、磨损、偏心、轴承内圈、外圈、滚动体损坏等故障问题进行综合分析研究,并采用智能分类算法支持向量机(Support Vector Machines)对风电机组的齿轮箱轴承进行故障分析和分类研究,采用实测实验模拟故障数据验证故障诊断算法的可行性和准确性,为风力发电机组故障诊断提供一种新的解决方法。研究内容和结论如下： (1)从机械故障诊断的基本原理出发分析研究风力发电机组齿轮箱滚动轴承机械振动故障机理,分析出各个部件出现故障的特征频率,明确不同部件故障所对应的故障特征,为后续故障诊断和分类实验验证提供数据支撑和理论依据。 (2)在获得故障测试数据之后,采用小波分解与EMD分解阈值方法进行降噪处理、频谱和包络谱分析,观察分析频谱图中的故障特征量,提出了改进EMD阈值降噪方法,并验证其可行性和优越性 (3)在故障数据中选取峰-峰值、有效值、方差和峭度值指标作为故障特征量,采用支持向量机分类识别算法对所选取的故障特征量组成的训练样本进行训练,构成故障诊断基本模型,然后采用网格搜索方法、遗传算法、粒子群算法这三种参数优化算法对支持向量机诊断模型的参数进行优化,以获取齿轮箱滚动轴承的故障点的精确定位和故障类型的有效辨识,仿真结果表明,网格搜索法虽然计算速度相对较快一些,但是故障类型分类准确率较低。遗传算法容易陷入局部最优并且计算速度相对较慢,分类效果欠佳。粒子群优化算法分类准确率最高,计算速度比遗传算法快,但是收敛性差。
[Abstract]:The gearbox is an important part of the transmission system of wind turbine. The rolling bearing of the gearbox is one of the parts with high failure rate in the transmission chain. Bearing failure not only affects the normal operation of the unit, but also affects the safe and stable operation of the power supply network. Therefore, it has important practical significance and practical value to diagnose the bearing fault of wind turbine gearbox quickly. In this paper, the basic structure and failure mechanism of wind turbine are analyzed. Based on the fault features and their characteristic frequencies, the improved threshold method of empirical mode decomposition (Empirical Mode Decomposition) is used to pre-process the vibration signals of rolling bearings in the event of failure, then the fault features are extracted and the fault types are analyzed. Combining time and frequency domain analysis technology with the knowledge of rotating machinery fault theory, this paper aims at the problems of common mechanical faults in the mechanical transmission system of wind turbine, such as tooth breaking, pitting, abrasion, eccentricity, bearing inner ring, outer ring, etc. The problems of rolling body damage and other faults are comprehensively analyzed and studied. The intelligent classification algorithm, support vector machine (Support Vector Machines), is used to analyze and classify the gearbox bearing of wind turbine. The feasibility and accuracy of the fault diagnosis algorithm are verified by the simulated fault data of the measured experiments, which provides a new method for the fault diagnosis of wind turbine generator. The research contents and conclusions are as follows: (1) based on the basic principle of mechanical fault diagnosis, the mechanism of mechanical vibration failure of roller bearing of wind turbine gearbox is analyzed, and the characteristic frequency of each component fault is analyzed. The corresponding fault characteristics of different components are defined, which provides data support and theoretical basis for subsequent fault diagnosis and classification experiment verification. (2) after obtaining the fault test data, wavelet decomposition and EMD decomposition threshold method are used to reduce the noise, the spectrum and envelope spectrum are analyzed, the fault characteristic quantity in the spectrum chart is observed and analyzed, and the improved EMD threshold denoising method is put forward. The feasibility and superiority of the method are verified. (3) the peak-peak value, effective value, variance and kurtosis are selected as the fault characteristic variables in the fault data. The support vector machine (SVM) classification and recognition algorithm is used to train the training samples composed of the selected fault features to form the basic fault diagnosis model. Then the grid search method and genetic algorithm are used. Three parameter optimization algorithms, particle swarm optimization (PSO), are used to optimize the parameters of support vector machine (SVM) diagnosis model in order to obtain accurate location of fault points and effective identification of fault types of gearbox rolling bearings. The simulation results show that, Although the computing speed of grid search method is relatively fast, the accuracy of fault classification is low. Genetic algorithm (GA) is easy to fall into local optimum, and the computation speed is relatively slow, and the classification effect is not good. Particle swarm optimization (PSO) has the highest classification accuracy, faster computation speed than genetic algorithm, but poor convergence.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM315

【参考文献】