摘要
SLM生产过程铺粉缺陷多采用人工间断性离线检测的方法,存在容易漏检、误检、效率较低和响应时间较长等问题,极大地影响了生产效率和生产品质。因此,搭建了一种基于图像处理的SLM铺粉缺陷检测识别与分类系统。该系统包括图像采集及缺陷提取、缺陷定量表征与分类识别两部分。在图像采集及缺陷提取部分,搭建了铺粉缺陷检测硬件系统,提出了基于概率快速自动选取阈值、滞后阈值分割等方法用于分割缺陷;在缺陷定量表征与分类识别部分,提出了缺陷定量表征与CART决策树方法进行缺陷种类的细分。结果表明,本系统对铺粉缺陷的检测识别与分类精度可达到96.67%,响应时间在400 ms左右,具有较高的识别精度与检测效率,显著提升了生产效率和生产品质。
Artificial intermittent offline method is commonly used during SLM powder spreading process, which exists in problems such as missing, false detection, low efficiency, long response time, greatly affecting the production efficiency and quality. On this basis, a defect recognition and classification system of SLM powder spreading was established based on the image processing, which included image acquisition and defect extraction as well as defect quantitative characterization and classification recognition. For the former one, a hardware detection system of powder spreading defect was built up, and methods for defect segmentation such as fast automatic threshold selection and lag threshold segmentation based on probability were proposed. For the latter one, defect quantitative characterization and CART decision tree method were proposed to subdivide defect types. The results indicate that the detection, identification and classification accuracy of the system can reach 96.67%, and the response time is about 400 ms, which exhibits high identification accuracy and detection efficiency, greatly improving the production efficiency and quality.
作者
伍缘杰
徐晓静
计效园
杨欢庆
彭东剑
武博
周建新
殷亚军
沈旭
Wu Yuanjie;Xu Xiaojing;Ji Xiaoyuan;Yang Huanqing;Peng Dongjian;Wu Bo;Zhou Jianxin;YinYajun';Shen Xu(State Key Laboratory of Material Processing and Die&Mould Technology,Huazhong University of Science and Technology;Xi'an Aerospace Engine Co.,Ltd.)
出处
《特种铸造及有色合金》
CAS
北大核心
2023年第1期23-28,共6页
Special Casting & Nonferrous Alloys
基金
国家自然科学基金资助项目(51905188,52090042,52275337)
KGW资助项目(2019XXX.XX4007Tm)。
关键词
SLM
铺粉过程
缺陷检测
缺陷分类
图像处理
SLM
Powder Spreading
Defect Detection
Defect Classification
Image Processing
