I have acquired several development kits containing eight BME688 sensors. I did several experiments, collected data with the kits at the same time with identical conditions then exported to BME AI-Studio and trained the data.
Training results seem pretty good and consistent but different from kit to kit.
I swapped only the sensor data from the kits that I have done in clean air and trained again and yes they are still different from each other.
I did 24 burn in process in clean air for all sensor kits at the same time.
Do the sensors have different characteristics or tolerances that could be causing this issue or any other cause?