首页 | 本学科首页   官方微博 | 高级检索  
     


Sparse Representations to Replace TOFD Images in Non-destructive Testing of Materials
Authors:Thouraya Merazi-Meksen  Akila Kemmouche  Malika Boudraa  Bachir Boudraa
Affiliation:1.Faculty of Electronics and Computer Science, University of Science and Technology Houari Boumediene (USTHB),Algiers,Algeria
Abstract:This paper describes a proposed method for the selection of relevant samples of ultrasonic signals during automatic material inspection. Instead of the well-known time of flight diffraction (TOFD) images, data are stored as a sparse matrix in which the elements only indicate whether a defect has been detected. This technique avoids storage of useless signals received during probe displacement in cases of low and high signal-to-noise ratios that correspond to coarse-grained and fine-grained materials, respectively. The approach is based on comparing the positions of maximum amplitudes, which are randomly located when signals only consist of noise but are in the same signal range when a defect is detected. The matrix elements are then applied as inputs to a self organizing map by neural networks to produce a normalized sparse matrix as the output, with a constant number of elements. This approach will be beneficial to enable the use of selected data in intelligent systems requiring a fixed number of inputs to characterize defects.
Keywords:
本文献已被 SpringerLink 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号