Theses and Dissertations - UTB/UTPA
Date of Award
Master of Science (MS)
Dr. Subhash Bose
Dr. Mounir Ben Ghalia
Dr. Immanuel Edinbarough
An automated robotic inspection system for electronic manufacturing has been developed to identify pin defects of IC packages mounted on printed circuit boards using surface mount technology. The automated robotic inspection system consists of two robots, a computer, a CCD camera with frame gabber for image acquisition, and a customized windows program using neural network for on-line defect identification. Gray scale images of the pins on IC packages are acquired using ambient light. The images are filtered and formatted to appropriate size, so that Matlab neural network tool could be used. The images are used to train neural networks using Matlab's Bayesian Regularization module. Optimal network was found to be a single-layer network with three outputs for each IC investigated. The weights and biases of each of the ICs investigated and the matrices of gray scale values for the IC images are saved as text files. A customized windows program uses these text files for on-line defect identification. The defect identification for the networks was found to be 100 percent for the two ICs investigated. The analysis and integration of an automated robotic inspection system for on-line monitoring of electronic manufacturing using neural networks is presented in this work.
University of Texas-Pan American
Copyright 2002 Roberto Alejandro Balderas.All Rights Reserved.