基于小波包的风力发电机组齿轮箱故障诊断研究

发布时间：2018-04-08 09:51

本文选题：风力发电机组　切入点：齿轮箱　出处：《新疆大学》2017年硕士论文

【摘要】：齿轮箱是风力发电机组的重要组成部件,也是故障发生率最高的部件,它的运行状态直接影响着风力发电机组的整体性能。因此开展风力发电机组齿轮箱的故障诊断对提高风能利用率和降低设备的维护成本等方面具有重要的现实意义。本文以双馈型异步风力发电机组齿轮箱为研究对象进行故障诊断研究。首先简述了风力发电机主要的内部构造以及能量转换过程,重点分析了齿轮箱中齿轮与轴承的故障类型及其振动原理。其次,选用小波包变换对故障振动信号进行信号消噪、小波包分解与重构,并与能量谱分析法相结合提取振动信号故障特征,组成故障特征向量。最后,将获得的特征向量分成训练集与学习集作为支持向量机的输入,进行故障识别。本文针对支持向量机的参数选择仍依赖于经验选取的情况,提出粒子群算法和遗传算法对支持向量机的参数进行寻优。采用凯斯西储大学试验台和实际风电场的轴承振动数据,对小波包和标准支持向量机以及参数寻优后的支持向量机相结合的故障诊断方法,进行仿真验证和实例分析,并做对比分析。仿真结果表明该方法能有效的提取故障特征,准确的进行故障识别,并在明显提高故障分类准确率的同时节省分类时间。因此该方法具有一定的实用性、推广性和准确性。
[Abstract]:Gearbox is an important component of wind turbine and also the component with the highest fault rate. Its running state directly affects the overall performance of wind turbine.Therefore, the fault diagnosis of wind turbine gearbox has important practical significance in improving the utilization rate of wind energy and reducing the maintenance cost of the equipment.In this paper, the fault diagnosis of doubly fed asynchronous wind turbine gearbox is studied.Firstly, the main internal structure and energy conversion process of wind turbine are briefly introduced, and the fault types and vibration principle of gear and bearing in gearbox are analyzed emphatically.Secondly, wavelet packet transform is used for signal de-noising, wavelet packet decomposition and reconstruction, and combining with energy spectrum analysis to extract the fault feature of vibration signal to form fault feature vector.Finally, the obtained feature vectors are divided into training set and learning set as input of support vector machine for fault identification.In this paper, particle swarm optimization (PSO) and genetic algorithm (GA) are proposed to optimize the parameters of SVM.By using the bearing vibration data of the test bench of case Western Reserve University and the actual wind farm, the fault diagnosis method combining wavelet packet with standard support vector machine and support vector machine after parameter optimization is simulated and analyzed.And make a comparative analysis.The simulation results show that this method can extract fault features effectively and accurately identify faults, and can obviously improve the accuracy of fault classification and save classification time.Therefore, this method has some practicability, generalization and accuracy.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】