This repository contains the documentation and core software logic for my Master's Thesis, developed in collaboration with Ferrari S.p.A. at the ALMA MATER STUDIORUM - UNIVERSITÀ DI BOLOGNA.
The central focus is a comprehensive lifetime model for SiC Power MOSFETs—the critical power switches—used inside high-performance automotive traction inverters. The work combines software modeling with detailed experimental characterization methods. It contains the comprehensive MATLAB implementation for predicting thermal fatigue lifetime in Power MOSFETs starting from its electrical characteristics and the actual speed / torque / current /voltage profile in time.
The automotive industry's push toward high-performance Electric Vehicles (EVs) demands smaller, lighter, and more powerful components. Silicon Carbide (SiC) technology is essential for this performance, but poses a major challenge: guaranteeing long-term reliability under extreme operating stress.
- Primary Goal: To create a predictive model that estimates exactly how fast these critical components will degrade under real-world driving conditions, ensuring vehicle safety and longevity.
-
Device Used: The model targets a high-power density, Dual-Side Cooled (DSC) SiC power module. We leveraged the module's internal body-diode for real-time Temperature Estimation (
$T_{vj}$ ). - Industry Standard: All modeling and proposed tests strictly follow the industry guideline ECPE AQG 324.
The main objectives were to create a robust digital tool for reliability engineering:
- Build a Digital Twin (MATLAB model) capable of predicting cumulative damage from any automotive mission profile (driving cycle).
- Define standard-compliant testing procedures (Power Cycling) to verify the model's accuracy against physical component failure.
This work was developed in collaboration with Ferrari S.p.A. Due to the sensitivity and intellectual property (IP) involved, the full model source code cannot be shared. This repository provides the full thesis documentation and key code examples illustrating the model's core logic.
-
High-Fidelity Electro-Thermal Modeling:
- We used Electro-Thermal Modeling and Thermal Impedance (
$Z_{th}$ ) networks to precisely calculate the virtual junction temperature ($T_{vj}$ ), which dictates component failure. This calculation of$T_{vj}$ is the single most critical step for lifetime prediction.
- We used Electro-Thermal Modeling and Thermal Impedance (
-
Stress Counting Logic:
- The model relies on the Rainflow Algorithm—the industry standard for fatigue analysis—to process complex, noisy temperature data. This algorithm transforms hours of driving into a quantified set of specific thermal stress cycles that cause damage.
-
Damage Quantification:
- The model uses Miner’s Rule and the Norris & Landzberg Acceleration Factor (AF) to convert counted stress cycles into a single, cumulative measure of remaining component life.
-
EOL Criteria & Testing:
- Failure is defined by thresholds based on industry standards (e.g., +5% drift in
$V_{DS,on}$ ). We propose a synchronized test bench utilizing a Source Measurement Unit (SMU) for microsecond-accurate Temperature Estimation during accelerated stress tests.
- Failure is defined by thresholds based on industry standards (e.g., +5% drift in
This thesis successfully formulates a robust, standards-compliant tool for linking real driving conditions to component failure. The developed MATLAB model ensures the long-term reliability, safety, and system integrity of next-generation high-performance EV powertrain systems.
This repository is licensed under the MIT License.
You can access the full thesis in PDF format here.
For any questions or further information, please contact me at tmatteos@gmail.com.