a) WHO AM I
As professional electrical engineer for 25 years and active element14 member, I have learnt a lot here and I am really willing to share with something different with other members.
I am fascinated by deep learning since last year and posted blog on BeagleBone AI - Review .
But deep learning is not omnipotent and can not solve everything. Sometimes the task is too tough to be completed, especially with limited computation resources within limited times.
In other word, I may fail to complete this experiment as I have envisioned. But I can ensure something different in my final blog, which may twinkle your imagination.
(b) WHAT TO DO with vibration sensors
To trace the healthy of large motors which is hard to maintenance in static state. Similar scenary as of gas-driven thermal engine on cars can be good alternative.
The Kemet Vibration Sensor with 20Hz to 15kHz can fixed near the engine of car, to show different mode of the car, such as braking, bumping, etc.
(c) HOW TO DO with the experiment
Step 1,Build one portable tool for vibration tracing on car, recording the data from Kemet vibration sensor
Step 2, Build one CNN model and train with DL framework. hopefully, the model can be convergence to one valid result.
Step 3, Deploy the model on the Embedded platform to be used as predictive maintenance hub.
The final blog is comprised of three mini-experiments. I would like to include,
- GUI tools touchGFX or some lightvgl
- one new Deep Learning framework,
- one new Development Board from ST, STM32H747-DISCO , so as to deploy with embedded xcube-AI or some new board like STM32G431-nucleo
(d) Kemet Vibration Sensors, Kemet Vibration Sensors, Kemet Vibration Sensors,
What ever software of hardware to be used, lies in center is Kemet Vibration Sensor, try to experiment as much as I can to understand fully of the parameters and performance.