Glue level estimation through automatic visual inspection in PCB manufacturing

Abstract

Nowadays, the increasing use of automatic visual inspection approaches in the manufacturing process is remarkable. The automation of production lines implies profitability and product quality. Moreover, optimized human resources result in process optimization and production increase. This work addresses the problem of optimizing the glue tube replacement in Printed Circuit Boards (PCB) manufacturing, warning a human operator only just in time to replace the glue tube. We propose an approach to estimate the glue level, in the glue injection process, during PCB manufacturing. The proposed methodology is composed of three main steps: i) Pre-Processing; ii) Feature extraction; and iii) Glue level estimation through machine learning. The proposed predictive model learns the relation between visual features and the glue level in the tube. Real and simulated experiments were carried out to validate the proposed approach. Results show the obtained Root Mean Square Error (RMSE) measure of 0.88, using Random Forest regression model. Furthermore, the proposed approach presents high accuracy even regarding noisy images, resulting in RMSE measures of 3.64 and 4.15 for a Salt and Pepper and Gaussian noise, respectively. Experiments demonstrate reliability and robustness, optimizing the manufacturing.