The prediction of electricity values in the process of electrolytic copper plating in PCB production through machine learning

การพยากรณ์ค่ากระแสไฟฟ้าในกระบวนการชุบทองแดงด้วยกระแสไฟฟ้าในการผลิตแผ่น PCB โดยใช้การเรียนรู้ของเครื่อง

  • Wikanda Phaphan Department of Applied Statistics, Faculty of Applied Science, King Mongkut’s University of Technology North Bangkok
  • Benjaporn Pratheeparunothai
  • Chanaphun Chananet
Keywords: Electrolytic copper plating, Support vector regression, Multiple linear regression, Decision tree regression

Abstract

The purpose of this study was to predict electricity values using in the process of electrolytic copper plating to increase the thickness of copper in holes and on the surface of the PCB board according to customer’s needs. The configuration data of electricity, collected from an electronics components manufacturing company, were the data that went back from January 2018 to September 2019. The studied data were divided into two data sets (according to types of machines using copper plating): line gate type and line VCP. We prepared the proper data to save it into the database by employing the best prediction model for each type of machine and using the R programming language for analyzing and predicting. The results showed that the lowest MSE and MAPE values which derived from the line gate of the support vector regression model are 1.6259 and 5.1991, respectively, followed by the multiple linear regression and the decision tree regression model. The MSE and MAPE values of the line VCP of the support vector regression model are also minimal values equal to 5.5466 and 7.8465, respectively, followed by the multiple linear regression and the decision tree regression model. Therefore, we use the support vector regression model to predict electricity values in the process of electrolytic copper plating for two types of machines to increase the accuracy in assigning electricity values and reduce working time due to an error in controlling electric current.

Published
2020-06-28