Power converter fault diagnosis using machine learning techniques

Abstract

Semiconductor devices are used in various power converters. These devices are often exposed to particularly tough operating conditions; they must withstand large amount of power with frequent fluctuations. Being the most vulnerable component in a power converter, these devices are bound to have high failure rates. This has led to an increase need for maintenances and repairs, therefore, increasing the cost of energy conversion. Reliability research in power electronics has been carried out for decades and is now moving from solely statistical approach to more physics-based approach. Temperature has also been cited as having the most significant impact on reliability of power electronics. Therefore, electrical-thermal analysis and simulation are necessary to perform reliability research.