Currently the best way to tackle sensor aging is by including data from both boards(old and new sensors) for training.
So, if i didn't misunderstand what you are saying, this means that even if someone samples data with the dev kit and then proceeds to train an algorithm, the resulting configuration works only with the devkit and not with other sensors connected to other MCUs?
Currently, you record data with dev kit, the sensor on dev kit run with the configuration file generated by BME AI-Studio will have better performance than other sensors.
Looks like the training is not working at all on a single sensor. I was collecting the data from the single sensor getting the samples for about half of hour for normal area condition, after that with cofee and another time, as it was described in the video. After the raw data have been collected i have imported this file to Bosh AI studio, but looks like the cicles are considered only for the first speciment and the rest is with 0... for the first one it is 48 even though i was more than 1000 sampling collected for 2 hours. Is there way to configure the AI for one sensor or there is a limitation on it to be working only with your dev tool?
The BME AI-Studio itself does not limit how many sensors are used to collect data, and when it is collected, as long as the imported data format meets the requirements. Detailed data format requirement, reference applicatoin note: https://www.bosch-sensortec.com/media/boschsensortec/downloads/application_notes_1/bst-bme688-an001....