Modeling component placement errors in surface mount technology using neural networks

Although today's electronics assembly equipment is able to generate huge volumes of computer-integrated manufacturing (CIM) data for efficient identification of machine maintenance requirements, most of this information is currently underutilized on the manufacturing floor. This paper presents a strategy to address this situation through the development of an intelligent machine maintenance tool (IMMT), a software-based maintenance system designed to analyze available equipment data in order to intelligently schedule both preventive and corrective maintenance operations. IMMT uses neural network based modeling techniques to identify when the collected tool data indicates maintenance actions are warranted. This paper provides an initial investigation of the applicability of these methods to model the frequency of nozzle and feeder errors in a surface mount component placement system. The distribution of component errors is modeled using back-propagation neural networks to estimate the amount of time until the next error occurs. Predictions for each nozzle and feeder in the system are then compared, allowing the preventive maintenance system to provide a ranked list of potentially faulty components. IMMT will ultimately be an on-line system for data-driven (as opposed to fixed-interval) preventive maintenance scheduling and dynamic diagnosis of machine faults.