Fault Detection of Air Defense Radar Systems Using Machine Learning

Hyungkwon Lee, Chulgyu Francis Kim, Min-Seop Kim, Young-Ho Kim, Han-Kyoung Park, Jong-Seok Lee

Abstract: This research presents an application of machine learning for condition-based maintenance of air defense radar systems. We established a data acquisition system during the evaluation of an air defense system. Data was collected in both faulty and normal operational states, enabling the development of classification models capable of distinguishing between these states. To address the class-imbalance issue, various resampling techniques were applied. We assessed the potential of effectively discriminating malfunctions from normal conditions and explored key factors influencing this discrimination. Our analysis results demonstrated the feasibility of accurately identifying malfunctions, achieving a true positive rate of 0.840 while maintaining a false discovery rate of 0.495.