SIA OpenIR  > 智能检测与装备研究室
轴承柱状滚动体表面缺陷检测关键技术研究
Alternative TitleResearch on Key Technologies for Surface Defect Detection of Bearing Cylindrical Roller
张万智1,2
Department智能检测与装备研究室
Thesis Advisor杜劲松
Keyword圆柱滚子 缺陷检测 机器视觉
Pages75页
Degree Discipline检测技术与自动化装置
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract轴承零部件表面质量是衡量轴承产品性能的重要指标之一,其直接影响轴承在使用过程中的可靠性和稳定性,因此对轴承零部件表面进行检测具有重要意义。本文依托轴承零部件表面检测应用背景,开展了基于机器视觉的轴承柱状滚动体(圆柱滚子)表面缺陷检测共性关键技术研究。近年来,随着机器视觉相关理论和技术的不断发展,以机器视觉为基础的表面检测方法因其高效智能、非接触无损等优点在轴承圆柱滚子表面检测中得到了广泛的应用。由于轴承圆柱滚子表面具有反光度高、缺陷与背景对比弱、缺陷微小、曲线轮廓复杂且存在形位公差等特点,使得基于机器视觉的轴承圆柱滚子表面缺陷检测技术仍存在许多问题和难点:第一、在表面缺陷分割时,传统的阈值分割方法难以分割尺寸微小的表面缺陷;第二、在轮廓缺陷检测时,由于轮廓存在缺损和形位公差,使得难以有效建立标准轮廓匹配模板;第三、如何有效地对不同维度特征进行选择和组合,并克服单一缺陷分类识别模型的局限性也是表面缺陷分类识别中的难点。解决这些难点,并进一步提高圆柱滚子表面缺陷检测技术的综合性能,是本文首要研究问题。对此,本文主要进行以下几个方面的研究:在柱面缺陷检测过程中,本文基于遗传优化Otsu法提取滚子主体区域,提高了分割效率;针对柱面缺陷微小,难以检测分割这一特点,本文提出权重方差法,显著的提高了表面微小缺陷检测分割的精度。在端面缺陷检测过程中,本文以Tukey函数对最小二乘算法进行优化,降低了轮廓缺损点对轮廓标准圆拟合精度的影响,并设计了基于自匹配差分的轮廓缺陷检测方法;针对圆环表面缺陷检测,本文建立了基于统计阈值的缺陷判定准则。对于柱面缺陷分类识别,本文提出基于单变量特征评价和PCA降维融合的特征选择方法,并依据所选择特征建立了基于多模型集成的缺陷分类识别模型。为验证本文所设计轴承圆柱滚子表面缺陷检测相关算法的性能,以现有圆柱滚子样本进行实验。实验结果表明,本文缺陷检测算法能够准确检测出圆柱滚子柱面和端面存在的缺陷,缺陷检测精度为0.1mm;并且对柱面缺陷进行分类识别的结果能够满足现阶段的工业生产需求,算法整体具有较好的稳定性和鲁棒性。
Other AbstractThe product surface quality of bearing parts is one of the important indexes to measure the bearing performance. It directly affects the reliability and stability of the bearing during use. Therefore, it is of great significance to test the surface of bearing components. Based on the application background of bearing component surface inspection, this paper studies the key technology of common defect detection of bearing column rolling element (cylindrical roller) based on machine vision. In recent years, with the continuous development of machine vision related theory and technology, the machine vision-based surface inspection method has been widely used in the bearing cylindrical roller surface inspection due to its high efficiency, intelligence, non-contact loss and other advantages. Due to the high degree of glare on the surface of the cylindrical roller bearing, the weak contrast between the defect and the background, the small defect, the complex curve contour and the geometrical tolerance, the machine vision-based bearing cylindrical roller surface defect detection technology still has many problems and difficulties. First, in the case of surface defect segmentation, the traditional threshold segmentation method is difficult to segment small surface defects; secondly, in the profile defect detection, due to the contour defect and shape tolerance, it is difficult to effectively establish a standard contour matching template; Third, how to effectively select and combine different dimensional features and overcome the limitations of the single defect classification recognition model is a difficult point in the classification and identification of surface defects. How to solve these difficulties and further improve the comprehensive performance of cylindrical roller surface defect detection technology is the primary research issue of this paper. In this regard, this paper mainly studies the following aspects: In the process of cylinder defect detection, this paper extracts the roller body region based on genetic optimization Otsu method, which improves the segmentation efficiency. For the small cylindrical defects, it is difficult to detect the segmentation. This paper proposes the weight variance method, which significantly improves the surface microscopicity. The accuracy of defect detection segmentation. In the process of end face defect detection, the Tukey function is used to optimize the least squares algorithm, which reduces the influence of contour defect points on the accuracy of contour standard circle fitting, and designs the contour defect detection method based on self-matching difference. Surface defect detection, this paper establishes a defect determination criterion based on statistical threshold. For the classification and identification of cylindrical defects, this paper proposes a feature selection method based on univariate feature evaluation and PCA dimension reduction fusion, and based on the selected features, a defect classification recognition model based on multi-model integration is established. In order to verify the performance of the algorithm related to the detection of bearing cylindrical roller surface defects designed in this paper, experiments were carried out with existing cylindrical roller samples. The experimental results show that the defect detection algorithm can accurately detect the defects of cylindrical cylinders and end faces, and the defect detection accuracy is 0.1mm. The results of classification and identification of cylindrical defects can meet the current industrial production requirements. The overall has good stability and robustness.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25180
Collection智能检测与装备研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
张万智. 轴承柱状滚动体表面缺陷检测关键技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
轴承柱状滚动体表面缺陷检测关键技术研究.(3234KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张万智]'s Articles
Baidu academic
Similar articles in Baidu academic
[张万智]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张万智]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.