BACKPROPAGATION NEURAL NETWORK DALAM MENDETEKSI LABEL KOMPONEN KEPING PCB

Abstract

Production quality control in identifying defective components on the chip label Printed Circuit Board (PCB) is an integral part of the supervision of fabrication. It takes automatic inspection system for detecting and automated chip components on the PCB label in order to support the demands of the working world fast and accurate production. Lack of installation of components on the PCB due to the loss of chip components on the PCB label may result in reduced performance of the PCB chip. The purpose of this study is to prove the speed, accuracy and privilege Back Propagation Neural Networks in Automated Visual Inspection System on chip detection component on the PCB label. Labels are classified three types of resistor, capacitor and Elco (electrolit Condensator). In this study proposed a method of developing and building a prototype on-chip identification of PCB defects using image processing methods with the incorporation of backpropagation as a classification. Backpropagation training algorithm using the best validation performance is 1.5013e-010 at 150 epochs tested using the data obtained 10  tissue levels of accuracy around 98.34%.

 Keyword: Printed circuit board, automated visual inspection system, backpropagation, neural network.