Machine Learning-Based Analysis of Dynamic On-Resistance of GaN Power Semiconductors

As the use of power electronic devices has rapidly expanded, the adoption of Wide Band Gap (WBG) semiconductors, such as Silicon Carbide (SiC) and Gallium Nitride (GaN), has emerged to meet specific requirements, including operation at higher temperatures, voltages, and frequencies. Among these, GaN devices have gained popularity in fast-switching power electronic applications due to their ability to switch faster with lower power loss.

High-electron-mobility transistors (HEMTs) stand out as the most widely used GaN power electronic devices. However, they are susceptible to trapping phenomena. Traps exhibit strong time-dependent responses, leading to the dynamic on-resistance (R_on) of GaN semiconductors. The dynamic behavior of R_on can significantly impact the breakdown voltage and switching performance of GaN semiconductors.

The objective of this research is to investigate trapping phenomena and their consequences through the application of Machine Learning. Through a series of experiments, we aim to identify measurable features to develop a model capable of explaining and predicting the dynamic behavior of R_on and its confidence interval.

Assignment

• Comprehensive literature review.
• Experimental testing for generating a limited amount of data.
• Data analysis and model training.
• Thesis writing.

Duration: 9 to 12 months.

Location: ECTM and ESP lab.

Requirements

We are looking for a passionate Master student with the following requirements:

• A robust background in electronics, power electronics, or mechanical engineering.
• Proficiency in Machine Learning, with skills in both semi-supervised learning and unsupervised learning.
• A solid knowledge of data analysis, including graph theory, clustering, and Bayesian Experimental Design (BED).
• Familiarity with reliability analysis and health monitoring in the area of power semiconductors.

Contact

Last modified: 2023-12-04