Design and Characterization of a Fiber Bragg Grating-Based Sensor System for Multi-Parameter and Real-Time Monitoring

Abstract

Fiber Bragg grating (FBG) sensors are dielectric, chemically inert, immune to electromagnetic interference and intrinsically multiplexable but their application in multi-parameter monitoring is limited by the cross-sensitivity: one reflected peak cannot distinguish a temperature change from a strain change. In this paper, a four-element, wavelength-division-multiplexed FBG sensor system is designed and characterized which can measure, simultaneously and in real-time, the temperature and the axial strain from a single fiber, in addition to the thermal field at four points along an eight metre span. The architecture is modeled in OptiSystem 21: a broadband source, a three-port circulator, the bidirectional fiber span and four uniform gratings, positioned on a 2 nm grid at 4 different wavelengths of 1546, 1548, 1550 and 1552 nm. The reflected spectrum is then processed in a MATLAB co-simulation component that implements a thresholded centroid peak estimation and an inversion of a sensitivity matrix. The reflectivity of the modelled resonance is 59 % and the full width at half maximum is approximately 0.13 nm, which is 10 times narrower than the channel spacing. Single-parameter sweeps provide a coefficient of determination > 0.999 with a sensitivity of 10.0 pm/°C between 0 and 100 °C and 1.2 pm/µε between 0 and 1000 µε. A 2×2 sensitivity matrix with a determinant of −11 pm2/°C•µε and a full scale scaled condition number of 2.78 is obtained by a grating pair that is differentially packaged, with a strain sensitivity difference greater than an order of magnitude (1.2 pm/µε versus 0.1 pm/µε) and whose thermal sensitivities are matched. The inversion is applied simultaneously at seven distinct values of temperature and strain to reproduce the applied values of these measurands with an accuracy of 3 × 10⁻¹⁴ °C and 1.2 × 10⁻¹³ µε, which is limited by the arithmetic used and is independent of the accuracy of the method. Under noise, the centroid estimator finds the resonance at 1550 nm at 1549.9999 nm, and the demodulation error over a 1000 pm span is ±1.08 pm, which corresponds to 0.12 °C and 1.39 µε after inversion. Repeatability after five thermal cycles is 0.40 pm, thermal linearity is R² = 0.9998, the hysteresis is 92.69 pm (full loop 0–100 °C) and the first order time constant of the interrogation chain is 100 ms sampled at 20 Hz.