Enable MOSFET junction temperature prediction within 1°C RMS accuracy with edge AI technology
Operate power converter MOSFETs within rated junction temperature for safety and long-term reliablity using edge AI models
Application overview
Operating the MOSFETs within rated temperature range is critical for safe and reliable operation of power converter system. However, fast transients, abnormal load conditions, can cause transient junction temperature swings that may exceed the rating and are challenging to measure using convetional temperature measurement techniques such as on-board sensors, thermcouples, etc.
Edge-AI based temperature prediction solution can improve the accuracy of junction temperature measurements using multi-dimensional inputs such as operating current, voltage, coolant flow rate and the gross temperature reading from the sensor.
Our innovative approach leverages a multi-layer perceptron edge-AI model trained on a proprietary temperature measurement dataset, running seamlessly on the TI's C2000 MCUs. This technology enables more accurate junction temperature monitoring.
Starting evaluation
Data collection
To obtain accurate reference data for switch temperature:
1) Direct temperature measurement
- Mount a thermocouple or similar temperature sensor directly on the power MOSFETThis provides the "golden reference" temperature values
2) Comprehensive operating conditions, operate the power converter across various conditions:
- Different input voltages
- Varying load levels
- Multiple cooling conditions
3) Collect transient events
- Transient conditions cause the greatest error in conventional temperature estimation methods
- Training data must include rich representation of transient events to develop accurate edge AI models
Available resources:
- Proprietary dataset generated using the above methodology for model training and evaluation. Use our provided dataset or bring your own collected data
Data quality assessment
The golden reference dataset is obtained by:
Mounting a temperature sensor as close as possible to the MOSFET junction, typically connecting the temperature sensor directly to the MOSFET package. This provides the closest approximation to actual junction temperature
Figure 1.1 -Temperature prediction MOSFET
Figure 1.2 - Temperature sense data - good
Figure 1.1 - Predicted MOSFET case temperature
Build and train your model
With your dataset ready, explore, train, and evaluate models using the TI tinyML Model Zoo
Find the right model for your needs
Access our library of optimized generic time-series models, scalable across performance and power requirements.
For MOSFET Temperature prediction , refer to the model in the TI tinyML Model Zoo under examples → mosfet_temp_prediction to achieve highly accurate temperature prediction.
Deploying your model
Start with our command line tools, our tinyml-modelzoo hosts a more extensive and flexible set of model-development tools, including Bring your own data (BYOD) or Bring your own model (BYOM)
Choosing the right device for you
The C2000 MCU family delivers scalable performance for your broad AI need, including C29's Very Long Instruction Word (VLIW) parallel performance with F29H859TU-Q1, TinyEngine™ NPU with TMS320F28P559SJ-Q1 and many others.
All the hardware, software and resources you’ll need to get started
Hardware
LAUNCHXL-F29H85X
C2000™ real-time MCU F29H85x LaunchPad™ development kit enable rapid development of Edge AI use case.
F29H85X-SOM-EVM
F29H85x controlSOM evaluation module enable rapid development of Edge AI use case.
LAUNCHXL-F28P55X
C2000™ real-time MCU F28P55X LaunchPad™ development kit enable rapid development of Edge AI use case.
Software & development tools
CCStudio™ Edge AI Studio
Comprehensive set of tools for training, compiling and deploying a model to TI edge AI devices. A model selection tool is available to view pre-generated benchmarks of popular models.
CLI Tools
A command line interface for advanced users, who want to develop their own model. Use this end-to-end model development tool that contains dataset handling, model training and compilation.
F29-SDK
Foundational Software Development Kit (SDK) for F29 real-time MCUs.
Supporting resources
The starting guide shows developers which aspects they need to consider for building an Edge AI application on TI Processors, and the tools they may require to do so