无损检测和模式识别技术在粉末注射成形中的应用  

APPLICATION OF NONDESTRUCTIVE EXAMINATION AND PATTERN RECOGNITION TO POWDER INJECTION MOLDING

作　　者：李振刚[1] 何新波[1] 曲选辉[1] 班晓娟[2] 周瑜[2] 

机构地区：[1]北京科技大学材料科学与工程学院,北京100083 [2]北京科技大学信息工程学院,北京100083

出　　处：《粉末冶金工业》2008年第3期27-30,共4页Powder Metallurgy Industry

摘　　要：从粉末注射成形技术智能化的角度出发,将无损检测技术与模式识别技术相结合,提出了一种高效可行的缺陷检测诊断方法。本文通过工业CT检测注射成形坯中的两种常见的缺陷,在对缺陷图像进行图像处理后进行傅立叶变换,选取缺陷图像的傅立叶变换作为特征参数来进行BP神经网络的输入后,获得了较好的识别效果。将识别结果输入专家知识库后给出相应的产生原因和解决办法,为实现缺陷的智能控制提供了一种有效的解决办法。In terms of intelligent powder injection molding, an efficient and feasible method combining nondestructive examination （NDE） and pattern recognition （PR） is proposed , Two typical defects in powder injection molded semi-finish products were studied. First,industrial computation tomography was used to identify the defects and obtain the images. Then the images were processed by image processing and Fourier conversion. The results of Fourier conversion were used as characteristic parameters to input into the BP neural network. At end, good recognition results were obtained. The reasons and corresponding solutions were obtained when the recognition results were input into expert knowledge database. This method is effective to control defects intelligently.

关 键 词：注射成形 无损检测 模式识别 

分 类 号：TF124.39[冶金工程—粉末冶金]