Methods for on-board Condition Monitoring of SiC MOSFET Based Converters




Journal Title

Journal ISSN

Volume Title



The power electronics industry is continuously striving to improve the efficiency and density of power converters. At the same time, with increasing electrification and automation across application domains, the power electronic systems are expected to meet stringent reliability requirements, especially in safety-critical applications such as aerospace, autonomous vehicles, data centers, etc. Silicon Carbide (SiC) power semiconductor devices promise significantly superior electro-thermal performance to traditional silicon IGBTs and MOSFETs. However, given their relative nascence, the field reliability of SiC devices is unproven and certain fundamental reliability challenges exist. This dissertation aims to study on-board condition monitoring methods as a potential solution to addressing reliability challenges with SiC MOSFET based converters. The dissertation first presents a detailed architecture for a modular, highly-scalable accelerated testing platform for SiC MOSFETs. The proposed testing setup enables rapid aging of large batches of SiC MOSFETs for the purpose of generating large datasets to study long-term reliability, and identify electrical precursors that can be used for on-board condition monitoring of SiC devices. Testing on a batch of discrete SiC MOSFETs using the developed test platform revealed the frequent occurrences of gate-open failure in discrete SiC MOSFETs. Therefore, in this dissertation, gate-open failures are systematically studied in the context of SiC MOSFETs, and potential causes for SiC MOSFETs’ increased susceptibility to gate-open failures is discussed. Importantly, a robust cycle-by-cycle gate-open failure detection solution is presented and its superior performance over traditional protection schemes is experimentally validated. Lastly, this dissertation proposes an end-to-end practical online condition monitoring solution for SiC MOSFET- based traction inverters using device on-state resistance (Rds−on) as an aging precursor. The proposed solution includes accurate on-board on-state resistance (Rds−on) measurement circuits along with code-efficient data acquisition and filtering algorithms. Importantly, the presented solution uses a stochastic Bayesian state-of-health estimation algorithm. The algorithm presents an elegant solution to the fundamental problem of separating aging-related Rds−on change from operating conditions-related changes by exploiting the symmetrical nature of the inverter’s operation. In particular, the presented solution is highly scalable as it automatically accounts for device and system level variations and eliminates the need for extensive system/device specific calibration.



Engineering, Electronics and Electrical