Customer Needs
Enhance nominal performance of traction inverters.
Reduce the bill of materials.
Challenges
Existing thermal management systems operated without real-time insight, requiring oversized safety margins. This led to reduced efficiency, performance, and profitability.
Solution
VTS is a software solution embedded into control units, offering real-time temperature predictions without hardware sensor limitations.
It predicts outcomes to optimize control decisions, enhancing traditional hardware sensors' capabilities.
Technology and methodology
Model Order Reduction (MOR): Converts detailed finite element models (1 million DOFs) to 24 DOFs for real-time use.
Embedded system requirements:
RAM usage: < 5 kB
Flash size: < 100 kB
Execution time: 36 µs on Aurix TC3 processors.
Accuracy: Temperature predictions with deviations < ±5°C under all conditions.
Key results
Reduced safety margins from 25°C to 5°C.
Increased nominal current by 10%.
Improved peak current duration by 4x.
Enhanced junction condition monitoring to improve the lifetime of the device
Diode and IGBT temperature predictions demonstrated high accuracy with minimal error (mean error ~1.5°C to 2.5°C depending on lifecycle stage).
The reliability of VTS was validated with consistent accuracy across different operational states and conditions.
Tool
Twin Fabrica: A software engineering platform for creating, simulating, and implementing virtual sensors in an integrated and scalable way.