INTEGRATED NEURAL-NETWORK MODELING FOR ELECTRONIC MANUFACTURING

This paper addresses issues involved in the modeling of electronic man ufacturing processes for optimization and control using artificial neu ral networks (ANNs). A modeling methodology is presented which integra tes a number of techniques to counter the commonly experienced problem s of selecting the 'right' network structure, over-training and long t raining times in building economical and accurate ANN models. This met hodology has been implemented as an automated user-friendly ANN modeli ng software CU-ANN. The main features of our methodology are data pre- processing, 'simple to complex' network structure approach and simulta neous training and testing. The neural networks considered have feed f orward architecture and use error back-propagation algorithm for train ing. We have successfully applied this ANN modeling methodology to a n umber of simulated and real-life electronic manufacturing problems. Th ese include stencil printing and simulated wafer fab. process data. Th e results indicate that our approach produces accurate, economical mod els and can handle a wide variety of data sets.