Supervisor:
Huai Wang

Co-Supervisor:

Assessment Committee:
Daniel Ioan Stroe (Chair)
Li Ran, University of Warwick, UK,
Vincent Heiries, CEA-LETI, France

Moderator:
Daniel Ioan Stroe

Abstract:

Three-phase inverters play a vital role in traction applications with a demand for zero-failure-period in the future, especially for safety-critical applications like automotive. However, subjected to prolonged operational and environmental stresses, the DC-link capacitor and power semiconductors in the inverter could be vulnerable and fragile, and consequently, unscheduled failure may occur. To reduce maintenance costs and enhance system availability, monitoring the operation and health status of these two reliability-critical components becomes essential. Existing condition monitoring solutions generally face challenges in terms of cost, complexity, and risk caused by additional hardware and software requirements. In addition, their accuracy is rarely verified under realistic operating conditions, and little effort has been devoted to further data analysis and algorithm implementation for decision-making. To address the above issues, a series of practical and effective condition and health monitoring schemes for traction inverters and their reliability-critical components are developed in this Ph.D. project.

Firstly, a special operation mode of traction inverters, the DC-link discharge after the system is shut down, is investigated, which is an inherent and frequent activity in traction applications like electric vehicles and is applied throughout the Ph.D. project to develop condition and health monitoring schemes. As the monitoring is implemented after the system shutdown, the potential risks and threats posed to normal system operation can be minimized.

Secondly, based on the discharging operation mode, an improved capacitance estimation method is developed for the DC-link capacitor in traction inverters, which is implemented with the existing information of the traction inverter without the need for additional hardware measurements or updates. By exploiting the discharge steady-state characteristics, the requirements in sampling rate, synchronization, and accuracy are greatly reduced, and a novel correction solution is proposed to eliminate and compensate for signal errors, ensuring the estimation accuracy in field applications. Compared with conventional estimation methods, this scheme outperforms in cost, complexity, and accuracy.

Thirdly, an innovative and complete condition and health monitoring scheme for power semiconductors in traction inverters is developed by using the DC-link discharge, encompassing data acquisition to further data processing for decision-making. A practical monitoring data acquisition approach is proposed, which addresses the challenge of obtaining device junction temperature in practice. In addition, by exploiting the low dynamics in the discharge, the measurement requirements for the health indicators and operating conditions of power semiconductors, including the sampling rate, synchronization, and accuracy, are greatly reduced, and the measurement disturbances from parasitic parameters are attenuated. Further, the data processing and calibration solutions for acquired historical data are provided, and two case studies are given demonstrating effective decision-making under realistic operating conditions, including the health assessment, remaining useful lifetime (RUL) prediction, and calibration for junction temperature estimation using thermo-sensitive electrical parameters (TSPEs).

Lastly, a novel and exceedingly simple health indicator, termed discharge time, is explored, which enables converter-level health monitoring for both the DC-link capacitor and power semiconductor devices in traction inverters. Its acquisition relies solely on the DC-link voltage information during the discharging operation mode, eliminating the need for additional software algorithms and hardware measurements. The relevant operating and environmental variables, as well as the further data calibration scheme, are given as well to facilitate the implementation of health monitoring. This monitoring scheme is expected to fundamentally address the cost, complexity, and risk concerns of prior arts.

Overall, the above studies contribute to the practical implementation of condition and health monitoring in real traction inverters, and proof-of-concept experiments have been conducted on both direct liquid-cooled and forced air-cooled inverter prototypes. The outcomes have been disseminated through a few journal articles and conference articles.