作者：南瑞亭1, 黄坚2
作者单位：1. 广州市交通技师学院，广东 广州 510540;
2. 华南理工大学机械与汽车工程学院，广东 广州 510640
关键词：微型直流电机;电机端盖;装配质量;深度学习;视觉检测
摘要：
针对目前微型直流电机端盖装配质量采用人工目视检测，存在主观程度高、信息化程度低的问题，该文提出基于区域推荐型卷积神经网络(R-CNN, region-convolutional neural networks)的微型直流电机端盖装配质量在线视觉检测技术。首先，应用Faster R-CNN目标检测方法，实现机壳冲压脚、正极、负极等端盖关键制造质量特征的识别与定位；根据电机型号对应的端盖装配质量需求，统计端盖上关键质量特征类型与数量(如机壳冲压脚及电机正极、负极、引线及插座等)，从零部件安装到位、冲压脚齐全、正极正确涂装三方面评价微型直流电机端盖装配质量。初步实验表明，该文方法可实现微型直流电机制造过程不同规格尺寸微型直流电机端盖装配质量视觉检测，单个电机检测时间不超过0.21 s，满足微型直流电机端盖装配质量在线视觉检测需求。
On-line visual inspection technology for the assembly quality of miniature direct current motor bearing support
NAN Ruiting1, HUANG Jian2
1. Guangzhou Communications Technician Institute, Guangzhou 510540, China;
2. School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, China
Abstract: For the current motor bearing support assembly quality using manual visual inspection, there is a high degree of subjectivity, low degree of information technology problems. In this paper, an online visual inspection technology based on region-convolutional neural networks (R-CNN) for the assembly quality of miniature direct current (DC) motor bearing support is proposed. Firstly, the Faster R-CNN object detection method is applied to identify and locate the key manufacturing quality features on the bearing support, such as the stamping feet, positive pole, negative pole, lead and socket, etc. The quality of the bearing support assembly of the miniature DC motor is evaluated in terms of the installation of parts in place, the complete stamping feet and the correct painting of the positive pole. The test proves that this method can achieve visual inspection of the assembly quality of miniature DC motor bearing support of different sizes during the manufacturing process of miniature DC motors, and the inspection time of a single motor does not exceed 0.21 s, meeting the demand for online visual inspection of the assembly quality of miniature DC motor bearing support.
Keywords: miniature direct current motor;bearing support;assembly quality;deep learning;visual inspection
2022, 48(3):124-128 收稿日期: 2021-08-05;收到修改稿日期: 2021-09-18
基金项目: 广东省重点领域研发计划项目(2019B010154003)；广东省质量技术监督局项目(2018CJ12)
作者简介: 南瑞亭（1981-），女，陕西西安市人，高级讲师，硕士，主要从事计量检测设备研发及相关专业教学工作
参考文献
[1] BHADORIYA A V S, GUPTA V K, MUKHERJEE S. Development of in-pipe inspection robot[J]. Materials Today-Proceedings, 2018, 5(9): 20769-20776
[2] LEE C, LIM M, KIM C, et al. Reliability analysis of accelerated destructive degradation testing data for Bi-functional DC motor systems[J]. Applied Sciences-Basel, 2021, 11(6): 1-9
[3] 王卫军, 徐友法, 赵鹏, 等. 分拣机器人在电机定子铁心生产线上的应用[J]. 微特电机, 2017, 45(9): 30-32+36
[4] 张明建, 曾伟明. 基于机器视觉的小型电机自动装配研究[J]. 制造技术与机床, 2017(6): 124-128
[5] 刘凯, 陈绪兵, 毛金城, 等. 基于机器视觉的电动机转轴轴承安装部位外径对比测量[J]. 轴承, 2019, 478(9): 49-52
[6] 王昕, 赵飞, 蒋佐富, 等. 迁移学习和卷积神经网络电力设备图像识别方法[J]. 中国测试, 2020, 46(5): 108-113
[7] 韩伟, 梁秋华, 段海峰, 等. 视觉控制在微型电机外壳冲压自动送料中的应用[J]. 锻压技术, 2020, 45(11): 162-167
[8] 彭伟康, 陈爱军, 吴东明, 等. 基于改进Faster R-CNN的水准泡缺陷检测方法[J]. 中国测试, 2021, 47(7): 6-12
[9] 杨健, 廖晨茜, 蔡晋辉, 等. 基于改进对抗迁移学习的滚动轴承故障诊断研究[J/OL]. 中国测试: 1-6[2021-08-29]. http://kns.cnki.net/kcms/detail/51.1714.tb.20210708.1412.034.html.
[10] LIU G X, HE B Y, LIU S Y A, et al. Chassis assembly detection and identification based on deep learning component instance segmentation[J]. Symmetry-Basel, 2019, 11(8): 1-12
[11] JIAN H, LIU G X, HE B Y. Fast semantic segmentation method for machine vision inspection based on a fewer-parameters atrous convolution neural network[J]. Plos One, 2021, 16(2): 1-13
[12] 刘桂雄, 何彬媛, 吴俊芳, 等. 面向视觉检测的深度学习图像分类网络及在零部件质量检测中应用[J]. 中国测试, 2019, 45(7): 1-10
[13] 刘桂雄, 刘思洋, 吴俊芳, 等. 基于深度学习的机器视觉目标检测算法及在票据检测中应用[J]. 中国测试, 2019, 45(5): 1-9
[14] 黄坚, 刘桂雄. 面向机器视觉检测的CNN语义分割方法进展[J]. 激光杂志, 2019, 40(5): 10-16
