A machine learning approach for simultaneous measurement of magnetic field position and intensity with fiber Bragg grating and magnetorheological fluid

This paper presents the simultaneous assessment of magnetic field intensity and position using a fiber Bragg grating (FBG) array immersed in magnetorheological (MR) fluid. The applied magnetic field leads to a variation of the MR fluid yield stress, which results in an axial strain on the FBG. As a well-known behavior of FBGs, the axial strain leads to a Bragg wavelength shift on the FBGs, which, in this case, is proportional to the magnetic field intensity and position. An array with 4 FBGs was used and characterized with respect to both magnetic field position and intensity. Then, a k-nearest neighbors' algorithm was proposed to classify the magnetic field position through the wavelength shift of the FBGs, where the magnetic field intensity is estimated from the FBG closest to the magnetic field position previously detected. Results show the feasibility of the proposed approach, where the algorithm accuracy is 100% for the best case and 86% for the worst case of magnetic field position, whereas a relative error lower than 5% was obtained on the magnetic field intensity estimation.