Fitting analysis and study on measured data of eddy current sensor via AutoML

Purpose This paper proposes a scheme using automated machine learning (AutoML) to address the nonlinear characteristics of eddy current sensors, making traditional linear fitting methods inadequate. The purpose of this paper is to enhance fitting accuracy through automated model selection and parameter optimization. Design/methodology/approach A measurement apparatus is used to collect displacement and oscillation frequency data from an eddy current sensor. AutoML technology is used to train models, predict data and generate fitting curves. The performance of the model trained using AutoML technology is evaluated against support vector regression (SVR), backpropagation (BP) neural network models and piecewise interpolation calibration methods. Findings The model trained using AutoML demonstrates superior fitting accuracy when compared to SVR, BP neural network models and piecewise interpolation calibration methods. It effectively reduces prediction output errors and automates model optimization, significantly improving the measurement precision of the eddy current sensor. Originality/value This study applies AutoML technology to the fitting analysis of eddy current sensors, automating the selection of optimal models and parameters. The approach overcomes the limitations of manual methods, offering enhanced accuracy and efficiency for nonlinear data fitting tasks.