Recurrent Neural Network-Based Stencil Cleaning Cycle Predictive Modeling

Abstract

This paper presents a real-time predictive approach to improve solder paste stencil printing cycle decision making process in surface mount assembly lines. Stencil cleaning is a critical process that influences the quality and efficiency of printing circuit board. Stencil cleaning operation depends on various process variables, such as printing speed, printing pressure, and aperture shape. The objective of this research is to help efficiently decide stencil printing cleaning cycle by applying data-driven predictive methods. To predict the printed circuit board quality level, a recurrent neural network (RNN) is applied to obtain the printing performance for the different cleaning aging. In the prediction model, not only the previous printing performance statuses are included, but also the printing settings are used to enhance the RNN learning. The model is tested using data collected from an actual solder paste stencil printing line. Based on the predicted printing performance level, the model can help automatically identify the possible cleaning cycle in practice. The results indicate that the proposed model architecture can predictively provide accurate solder paste printing process information to decision makers and increase the quality of the stencil printing process.