SiC EV Traction-Inverter Electrothermal Digital-Twin and FMANet Diagnostic Dataset

This archive provides the frozen research data, calibration objects, validation outputs, and trained networks supporting a four-class diagnostic study of a silicon-carbide electric-vehicle traction inverter. It contains scenario-level classical features and multirate time-series data for healthy operation, cooling degradation, gate-open faults, and on-state-resistance aging. The released populations comprise 999 development scenarios and 399 independent blind-test scenarios, together with six trained fusion and FMANet models and checksum manifests for integrity verification. Data were generated using a physics-based MATLAB/Simscape electrothermal digital twin; they are simulation outputs and are not hardware-in-the-loop or experimental measurements. The associated GitHub repository provides the model-building, data-generation, training, evaluation, and reproduction workflows.