Measurement Performance Improvement Method for Optically Pumped Magnetometer Based on Multilayer Perceptron

Optically pumped magnetometers (OPMs) are critical in magnetic field precision measurement applications. The core of OPMs is to use the proportion integration differentiation (PID) algorithm to maintain atoms in a magnetic resonance state. However, because the PID algorithm requires multiple iterations to stabilize the measurement results, the OPM cannot perform best when the magnetic field changes rapidly. To address this tackle, this article proposes a magnetic field measurement method based on the multilayer perceptron (MLP) model. By performing frequency modulation and synchronous demodulation on the RF field, the first-, second-, and third-order derivative values of the magnetic resonance curve are obtained. Using the mapping relationship between the three magnetic resonance derivative values and the detuning frequency and the MLP nonlinear fitting capability, a customized MLP model for calculating the magnetic field intensity is trained. The experimental results indicate that the proposed method can directly measure the magnetic field through a single measurement, improving the response speed over the PID method. The broader range of the higher-order derivative enhances the ability to handle rapid changes in magnetic fields. The sensitivity of the proposed method is 13fT/(Hz)(1/2). This research demonstrates the potential of integrating machine learning techniques with OPMs for precision magnetic field measurement scenarios.