基于谐波小波和支持向量机的风电叶片损伤识别研究

发布时间：2018-08-27 17:03

【摘要】：叶片是风力发电机的关键部件之一，对叶片损伤的研究越来越受到该领域研究人员的关注。由于叶片的结构巨大、形状不规则、材料铺层复杂并且长期工作在恶劣的环境下，所以当前需要解决的难题是如何实现叶片的健康监测。目前常用的监测手段是通过监测其模态来判断叶片的损伤状况，但该方法的缺点是敏感度低，而且一直未能得到有效地解决。针对这一问题，本文提出利用声发射技术对风电叶片损伤状况进行检测，并应用SVM(Support Vector Machine,支持向量机)对叶片的两类损伤模式进行识别。由于叶片在受到外力破坏时会引起材料内部应变从而产生声发射信号，通过对声发射信号进行采集和分析，能够实现对声发射信号源的识别。首先接通声发射传感器、信号放大器、数据采集卡和计算机等设备，搭建声发射信号采集实验平台，用耦合剂将声发射传感器固定在叶片上。然后人工对静态的单个叶片进行加载，模拟叶片的裂纹扩展和边缘破损两类损伤，并采集损伤时的声发射信号。采集到信号后，分别利用谐波小波包和db10小波包对声发射信号进行4层分解并计算信号的各频段能量值，，将所得能量值进行归一化处理后，所得数据作为特征向量，采用SVM对特征向量进行训练学习，建立叶片损伤识别模型。在进行叶片的损伤识别时，对两种小波包的特征提取效果进行了比较，仿真结果表明，采用谐波小波包和SVM结合的方法可以获得良好的识别效果。该方法能够有效地识别不同类型的损伤，有助于发现叶片初期损伤，使叶片可以得到及时地维护，防止损伤的进一步扩展。
[Abstract]:Blade is one of the key components of wind turbine. Because of the huge structure, irregular shape, complicated material layer and long term working environment, the problem that needs to be solved is how to realize the blade health monitoring. At present, the commonly used monitoring method is to judge the damage condition of the blade by monitoring its mode, but the disadvantage of this method is that the sensitivity is low, and it has not been effectively solved. In order to solve this problem, the acoustic emission technique is used to detect the damage of wind turbine blades, and SVM (Support Vector Machine, support vector machine (SVM) is applied to identify the two types of damage patterns. The acoustic emission signal can be obtained by collecting and analyzing the acoustic emission signal because the blade will cause internal strain of the material when it is damaged by external force, and the acoustic emission signal source can be recognized. First, the acoustic emission sensor, signal amplifier, data acquisition card and computer are connected to build the experimental platform of acoustic emission signal acquisition, and the acoustic emission sensor is fixed on the blade with coupling agent. Then, the static single blade is loaded manually to simulate the crack propagation and edge damage of the blade, and the acoustic emission signals are collected. After collecting the signal, the harmonic wavelet packet and the db10 wavelet packet are used to decompose the acoustic emission signal into four layers and calculate the energy values of each frequency band of the signal. After normalizing the energy value, the obtained data is used as the eigenvector. The feature vector is trained and studied by SVM, and the model of blade damage identification is established. In the process of blade damage identification, the feature extraction effects of two kinds of wavelet packets are compared. The simulation results show that the method of harmonic wavelet packet and SVM can obtain good recognition effect. This method can effectively identify different types of damage, help to detect the initial damage of leaves, enable the leaves to be maintained in time, and prevent the damage from spreading further.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM315

【参考文献】