基于特征综合提取方法的风电机组齿轮箱磨粒分类  

Wind turbine generator gear box grinding particle classification based on comprehensive feature extraction method

作　　者：苏连成[1] 张光远 SU Liancheng;ZHANG Guangyuan(School of Electrical Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《燕山大学学报》2020年第6期566-574,共9页Journal of Yanshan University

基　　金：河北省自然科学基金重点资助项目(F2018203256)。

摘　　要：针对当前对风力发电机组齿轮箱磨粒分类准确度不够高的问题,提出了一种基于磨损颗粒形状特征和边缘细节特征的磨粒特征综合选取方法。实验样本为风力发电机组齿轮箱润滑油液中提取的磨损颗粒,通过显微镜对磨粒图像进行采集。提取径向凹面偏差、面积偏差、细长程度、分形维数和曲率等相关磨粒边缘特征和形状特征作为磨粒分类的特征样本。使用随机森林算法对比了单独考虑形状特征或者单独考虑边缘特征时的分类结果,实验结果表明相比单独考虑形状特征或者边缘特征,使用本文提出的磨粒特征综合选取方法对磨粒进行分类的准确度更高,可以更准确地识别磨粒种类,为后续分析风电机组齿轮箱的运行状况提供了很大帮助。In this paper,a comprehensive selection method of abrasive grain features based on the wear grain shape features and edge detail features is proposed to solve the problem that the classification accuracy of the wear particles in the gearbox of the wind power generator is not high enough.The experimental sample is the wear particles extracted from the lubricating oil of the wind power generator,and the wear particle images are collected by the microscope.The edge features and shape features of the wear particles are extracted as the feature samples of the wear particles classification by extracting the radial concave deviation,the area deviation,the degree of slenderness,the fractal dimension and the curvature.The random forest algorithm is used to compare the classification results when the shape feature is considered alone or the edge feature is considered alone.The experimental results show that compared to considering the shape feature alone or considering the edge feature alone,the method proposed in this paper is more accurate to classify wear particles.This method also provides great help for subsequent analysis of the operation status of the wind power gearbox.

关 键 词：磨损颗粒 图像处理 凹面偏差 分形维数 颗粒分类 随机森林算法 

分 类 号：TG29[金属学及工艺—铸造]