An Artificial Neural Network to Support Package Classification for SMT Components

Abstract

A components package classification system (CPCS) based on an artificial neural network and feature selection of 2D patterns of electronic components is proposed to classify the package series of SMT components in this study. The accuracy of package classification will seriously influence the efficiency of the Design for manufacture (DFM) Check. The proposed CPCS can identify the 2D pattern of electronic components to classify the package series of SMT components. There are 19 features of the 2D pattern of electronic components to be used in this classification system. Through the experimental results, we got a 95.8% accuracy of classification. Some package series such as QFN and QFP were confused with each other as well as SOD and SODFL. CPCS can identify these kinds of package series by the specific feature selection. The experimental results show that CPCS can quickly and accurately classify the 2D patterns of electronic components to enhance the efficiency of DFM Check.