Machine Vision Inspection of Molds & Castings
机器视觉在线检测砂型与铸件
标签：机器视觉, AI检测, 在线质检, 传送带分选
在砂型线上以 2D/3D 相机做两道检查：合箱前扫描型腔（与 3D 模型比对，检出掉砂、冷铁位移与浇口破损，分辨率 0.1 mm/px）；落砂后铸件过传送带做表面缺陷 AI 分级（气孔、砂眼、冷隔），配深度相机测尺寸变形。某铜阀厂部署后漏检率降 60%，每班节省检验工 2 人。落地要点：频闪+偏振光源防反光、每类缺陷样本 ≥500 张、误报率压到 5% 以内，工人信任系统后才能替代目检。
Sand-casting lines deploy 2D/3D cameras at two stations: pre-close mold-cavity scans against the 3D model catch sand drops, shifted chills and broken gates (0.1 mm/px), and post-shake-out conveyors run AI surface grading (blowholes, sand inclusions, cold shuts) plus depth-camera dimensional checks. One bronze valve plant cut miss rates 60% and saved two inspectors per shift. Deployment keys: strobed polarized lighting against glare, 500+ samples per defect class, and false alarms below 5% — only then will operators trust the system over their eyes.
