Using machine learning algorithms to predict failure on the PCB surface under corrosive conditions

Abstract

A printed circuit board (PCB) surface can fail by corrosion due to various environmental factors. This paper focuses on machine learning (ML) techniques to build predictive models to forecast PCB surface failure due to electrochemical migration (ECM) and leakage current (LC) levels under corrosive conditions containing the combination of six critical factors. The modeling methodology in this paper used common supervised ML algorithms by accomplishing significant evaluation metrics to show the performance of each algorithm. The conclusion of this study presents that ML algorithms can create predictive models to forecast PCB failures and estimate LC values effectively and quickly. • Failure prediction on the PCB surface under many different corrosive conditions has been investigated. • Each corrosive condition is created by the combination of six critical factors, including pitch distance, contamination level, temperature, humidity, voltage, and contamination type, with three levels. • Standard supervised ML algorithms are used to train classifiers and regressors to predict failed or Not-failed conditions and LC values, respectively. • Common supervised ML algorithms have been used, including k-nearest neighbors, decision tree, random forest, support vector machines, and deep neural network.