Neural Network-Enabled Condition Monitoring of DC-Link Capacitors in Three-Phase Inverters

This work addresses condition monitoring of DC-link capacitors in three-phase inverters to ensure motor drive reliability. A neural network model is trained based on the DC-Link capacitor voltage data during a discharging operation mode, minimizing extra hardware needs. Experimental results on diverse operating conditions achieved 98% of accuracy, with 100% positive predictive value and 95% sensitivity, demonstrating robust classification and diagnostics across varying unseen operating conditions.