From Power Electronics to Data Intelligence: Comparative Study of SEPIC and ĆUK Converters with Weka Analysis

This paper presents a machine-learning-based approach for fault detection and classification in SEPIC and Ćuk DC–DC converters under fixed and variable input-voltage conditions. Four datasets were developed, including SEPIC with fixed Vin, SEPIC with variable Vin, Ćuk with fixed Vin, and Ćuk with variable Vin. Each dataset contains 200 instances representing four operating conditions: NoFault, CapacitorFault, InductorFault, and SwitchFault. Electrical parameters, including input voltage, output volt-age, input current, output current, and temperature, were used as classification features. Several machine-learning algorithms available in WEKA were evaluated, including Random Forest, J48, Random Tree, REPTree, Support Vector Machine, Multi-layer Perceptron, Naïve Bayes, and IBk. The results demonstrate that classification performance depends on both converter topology and input-voltage conditions. For the SEPIC converter, variable Vin generally improves classification accuracy, with Random Forest achieving the highest accuracy of 95.5%. In contrast, the Ćuk converter shows reduced performance under variable Vin. The findings demonstrate the potential of machine learning for automated fault diagnosis of power electronic converters under different operating conditions.