Development of a smart paint defects detection system

油漆缺陷表现为最佳油漆条件的不可避免的变化，例如粉末，工人错误，温度变化等，并表现为表面变化。有些缺陷很容易检测到，例如裂纹或划痕。否则，它们更难以检测，例如，粘附在部件上的小颗粒。这些缺陷导致零件不符合质量控制标准，从而导致零件拒收。熟练的操作员通常无法检测出80%的小缺陷。这些未检测到的缺陷构成了减少产品寿命的腐蚀源。 Rollstamp Mfg.，麦格纳外部公司的一个部门，是主要汽车公司的零件供应商。目前的汽车喷漆零件质量控制系统是基于对零件的100%人工目视检查。此外，手动过程不允许数据收集和快速反馈，而数据收集和快速反馈可用于识别问题的原因并相应地调整喷涂过程。该项目建立在与Rollstamp工程团队的协同关系以及通过NSERC/Engage和OCE/VIP赠款检测火山口油漆缺陷（行业确定的典型油漆缺陷之一）所获得的经验的基础上。 该项目旨在开发下一代智能生产级油漆缺陷检测系统Smart Inspector，该系统能够以可接受的精度和周期时间检测和分类不同零件几何形状的不同油漆缺陷。这将需要广泛的数据收集和使用机器学习策略来开发用于缺陷检测和分类的高级算法。该项目旨在协助行业伙伴努力提高其运营效率和产品质量，这对加拿大经济有直接好处。此外还将 提供培训机会，创造更高收入的工作，并因市场份额的增加而创造更多就业机会。

Paint defects appear as an inevitable variation of the optimum painting conditions, e.g. Powder, worker mistakes, temperature variation, etc. and show up as surface alterations. Some defects are easily detectable, e.g. cracks or scratches. Otherwise, they are more difficult to detect, e.g. Small particles bonded to the part. These defects lead to parts not meeting the quality control standards, resulting in the rejection of the part. Skilled operators usually do not detect 80% of minor defects. These non-detected defects constitute corrosion sources that reduce the life of products. Rollstamp Mfg., a division of Magna Exteriors Inc., is a part supplier for major automotive companies. The current quality control system for painted automotive parts is based on 100% manual visual inspection of parts. Additionally, the manual process does not allow for data collection and rapid feedback that can be used in identifying the cause of the problem and adjusting the painting process accordingly. The project builds on the synergetic relation with Rollstamp engineering team and the experience gained through an NSERC/Engage and OCE/VIP grant for detection of crater paint defect, one of typical paint defects identified by industry. The project aims to develop the next-generation smart production-grade paint defects detection system, the Smart Inspector, that is capable of detection and classification of different paint defects for different part geometries with acceptable accuracy and cycle time. Which will require extensive data collection and use of machine learning strategies to develop advanced algorithms for defects detection and classification. The project aims to assist the industry partner efforts to increase efficiency of its operation and quality of its products which is a direct benefit to the Canadian economy. In addition, it will provide training opportunities, create higher paying jobs and more employment opportunities due to increased market share.