Multi-dimensional fiber optic current sensor calibration method based on extreme gradient boosting

Fiber optic current sensors (FOCSs) are prone to environmental disturbances and have to be calibrated before going into service. A commonly adopted scheme is the single dimensional calibration method based on temperature. In this work, we propose a multi-dimensional FOCS calibration method based on the extreme gradient boosting (XGBoost) algorithm. Six operating parameters of the FOCS are chosen as the input of the calibration model, including the drive current of the light source, the average optical power of the photo detector (PD), the second harmonic of the PD output, the drive voltage of the piezoelectric transducer, the measured current, and the ambient temperature. The ratio error is acquired as the output. Temperature cycling experiments on three finished sensors in an environment of -40 degrees C to 72 degrees C are conducted. The experimental results show that the max absolute ratio error after multi-dimensional calibration based on XGBoost is 0.031% and the mean absolute ratio error is 0.003%, both of which are much lower than the single-dimensional calibration method based on temperature. It can be concluded that the proposed multi-dimensional calibration method based on XGBoost can effectively improve the accuracy and stableness of FOCSs. (c) 2024 Society of Photo-Optical Instrumentation Engineers (SPIE)