Dr Bing Ji, University of Leicester

This project is to demonstrate a new concept of the self-monitoring and resilient control (SMART) for Silicon Carbide (SiC) power devices, which represent a paradigm shift in automotive power electronics and underpins electric vehicle (EV) drivetrains. It will deliver highly-efficient, highly-reliable and high-power SiC MOSFET modules with through-life health monitoring and adaptive device-level control. It is anchored on the smart gate drive framework under developing in our lab.

It will address the following two key research challenges, each of which has associated objectives. (1) Conventional multichip power module approaches for mass-market automotive power electronics do not embrace the emerging opportunities offered by advanced microelectronics and multifunctional sensors and do not provide the flexibility and optimal customisation to serve a broad range of products or multiple inverter platforms. (2) Current automotive power electronics systems rely on reactive, post-fault protection and time-driven (periodical) maintenance with low effectiveness and resilience, i.e. short-circuit faults of traction inverters demands restoration time and may cause power interruption and catastrophic consequences. Field reliability is compromised by abnormal overstress events coupled with ageing and complex mission profiles.

Firstly, a key insight underpinning this research is to examine the power handling rating of SiC power modules with their wear-out failures, while considering the impact of different design and operational variations. Next, it is the first to use die-level sensing and distributed duty-ratio control for multichip SiC power modules to obtain online self-awareness and active thermal management capabilities. Finally, in-situ PHM is coordinated with edge computing and machine learning by focusing on data value exploitation. By capturing and extracting the specific features from data obtained from the design, testing, and usage phases, on both power module and inverter system levels, which is reinforced by the peer-to-peer comparison, a failure “fingerprint” will be made available to improve field robustness and resilience.