A Bayesian statistical method for large-scale calibration of MEMS sensors: a case study


The article  "A Bayesian statistical method for large-scale MEMS-based sensors calibration: case study on 100 digital accelerometers" authored by Andrea Prato, Francesca Pennecchi, Alessandro Schiavi (INRiM), and Gianfranco Genta (Politecnico di Torino), introduces an innovative approach to calibrating large batches of MEMS sensors.

The article suggests adopting statistical methods to replace, at least partially, the traditional sensor calibration procedures. Due to the high production volume, with millions of units produced weekly, the traditional one-by-one calibration approach becomes impractical in terms of time and costs.

The proposed Bayesian method involves experimentally calibrating a small sample of sensors, allowing the estimation of the number of reliable sensors in the entire batch and assigning an appropriate uncertainty to each sensor. This can be considered as a statistical calibration of the batch.

This approach reduces the number of experimental calibrations by incorporating information obtained from the previous calibration of a "benchmark" batch, performed "once and for all" and representing the entire production process.