Application of k-Nearest Neighbors Algorithm in Detection of Hidden Defects in Multilayer Printed Circuit Boards by Ultrasonic Method

This paper considers the application of ultrasonic method of detection of hidden defects of multilayer printed circuit boards of radio-electronic devices using the k-nearest neighbors algorithm in determining the technical condition of the studied samples of multilayer printed circuit boards. The method is based on the analysis of the spectral characteristics of the ultrasonic wave passing in the multilayer printed circuit boards volume from the radiation source to the receiver located in opposite parts of the multilayer printed circuit boards. Based on the results of the analysis and comparison of the spectral characteristics of the ultrasonic wave of the studied multilayer printed circuit boards sample and the reference model using machine learning using the k-nearest neighbors algorithm, a conclusion is made about the presence or absence of defects. The model of a serviceable multilayer printed circuit boards is used as a reference model, taking into account the scatter of geometric and physical-mechanical parameters of multilayer printed circuit boards.