文献类型：期刊文献

英文题名：Design and Realization of Pattern Classification Sorter for the Surface Defects in Hot-rolling Steel Plate

作者：Geng Rui-fang[1]; Cao Hui[1]; Ma Yong-hua[1]

机构：[1] Biochem. Eng. Coll., Beijing Union Univ., Beijing, China

第一机构：北京联合大学生物化学工程学院

年份：0

期号：25

起止页码：188-9, 131

外文期刊名：Microcomputer Information

收录：EI(收录号：12019961)

语种：英文

外文关键词：belts - cracks - hot rolling - learning (artificial intelligence) - metallurgy - neural nets - pattern classification - plates (structures) - sorting - steel

摘要：The hot-rolling steel plate is one of important products in metallurgy vocation,the computer automatic classified recognition system on the steel plate surface lacuna is the current research hot spot. The article mainly aimed at this application background to design one kind of sorter based on the BP algorithm artificial neural networks,so that to realized the classified recognition about the lacuna.Through training and the classified test on the neural network sorter, in the 10 testing datas,for 1,4,5 kind of flaws (corresponds belt-shaped flaw, concave-convex flaw and crack flaw separately),the distinguishing effect to be better,the accuracy rate of recognition is more than 90%;But this network to flaw type 2 (chap) and flaw type 3 (crowded irregular distribution flaw),the recognition has not achieved the anticipated effect, and also waits for further improves.