Machine Learning for In-Circuit Testing of Printed Circuit Board Assembly

Testing is an important procedure in a manufacturing process that leads to fabrication of high quality electronic components and modules. It can be facilitated through applying machine learning techniques and development of predictive and analytical models. The paper presents a method in support of test engineers at the In-Circuit testing of Printed Circuit Board Assembly when decision making has to be performed and testing problem has to be solved. Supervised machine learning algorithms: Support Vector Machine for resolving binary classification tasks and Random Forest for deciding the multi-class classification problem are utilized. The learners’ accuracy is evaluated and high results are achieved when 70% of the data set is used for training and 30% for testing. The accuracy of Support Vector Machine and Random Forest algorithms is compared to the accuracy of a deep learning algorithm. The proposed approach gives precise analysis and classification regarding the defects occurred during the mounting process on the Printed Circuit Board Assembly.