The simultaneous measurement of temperature, strain, and refractive index using a single optical fiber transducer remains a long-standing challenge because of the cross-sensitivity that arises when the three measurands induce overlapping spectral perturbations. We report a hybrid photonic-computational framework that combines a Ti3C2Tx MXene-functionalized tapered no-core fiber with a physics-informed neural network (PINN) and a Gaussian process regression (GPR) uncertainty model to address this limitation. The fiber transducer exploits the strong surface plasmon coupling enabled by the MXene over-layer on a 6 µm waist, producing a wavelength-shift sensitivity exceeding 2.1 nm/°C, 4.8 pm/µε, and 1860 nm/RIU in the 1.333–1.365 RIU range. The PINN embeds the cylindrical Helmholtz residual and the Drude permittivity of MXene directly into the loss function, yielding a data-efficient inverse model that requires only 1,400 labelled spectra yet generalizes across the full operating envelope. The GPR posterior calibrated the prediction intervals against measured spectra, achieving a 94.2 percent empirical coverage at the 95 percent nominal level. On five public open-access datasets (2024–2026) the proposed framework reduced the root-mean-square error by 3.4×, 2.9×, and 3.7 × for temperature, strain, and refractive index respectively when compared with a conventional matrix-inversion baseline, and by 1.7×, 1.6×, and 1.5 × against the best deep-learning baseline. These results demonstrate that physics-informed learning, combined with calibrated uncertainty, offers a viable route to deployable, trustworthy multiparameter fiber-optic sensing and aligns directly with the scope of the special issue on advanced fiber photonic devices based on two-dimensional materials and hybrid platforms.
Most existing electromagnetic and fiber Bragg grating (FBG) tilt sensors provide only single–axis measurements, limiting their capability for multidimensional structural assessment. In this study, an analytical tilt model for a sensitivity–enhanced equal–strength beam (SESM) was developed, and a circumferential biaxial FBG tilt sensor was fabricated for a measurement range of ± 30°. Calibration, spatial–tilt, and creep tests were conducted to evaluate its performance. The sensor achieved differential sensitivities of 81.234 pm/° in the yoz plane and 85.235 pm/° in the xoz plane, with R2 ≈ 0.9998 over the tested range. Under controlled room–temperature conditions (15) ± 0.5 °C), paired–FBG differential processing suppressed common–mode thermal drift. Spatial–tilt tests revealed weak but non–negligible cross–axis coupling, with coefficients of 4.76%, 3.17%, and 4.76% at fixed orthogonal–plane angles of 5°, 15°, and 25° in the fixed-xoz configuration (average ≈ 4.23%). During a 250–min creep test, the differential output remained approximately 1318 pm, indicating stable short–term behavior. The interrogator’s 1–pm wavelength–demodulation accuracy corresponds to angular–equivalent values of 0.0123° and 0.0117° in the yoz and xoz planes, respectively; these values are calculated from the interrogator specifications and do not represent independently validated system–level accuracy, which is bounded by the 0.05° accuracy of the reference platform. Compared with previously reported research prototypes, the proposed sensor provides a favorable combination of measurement range, differential sensitivity, and multidirectional measurement capability.
This study presents a numerical investigation of a novel star-shaped twin core photonic crystal fiber (TC-PCF) sensor for refractive index sensing of sucrose concentrations. The structural design and electromagnetic analysis are carried out using the finite element method implemented in COMSOL Multiphysics. Sensor performance is evaluated in the mid-infrared wavelength range, where key parameters such as confinement loss, wavelength sensitivity and amplitude sensitivity are extracted through MATLAB-based analysis. The proposed geometry incorporates a compact and well-defined micro-structured arrangement, which enhances the interaction between the guided modes and the analyte-filled region. Variations in sucrose concentration led to changes in refractive index, which modulates the modal properties and transmission characteristics of the fiber. The sensor achieves a maximum wavelength sensitivity of 8262.76 nm/RIU and an amplitude sensitivity of 3.3849 RIU−1, while maintaining negligible confinement loss. The combination of high sensitivity, low loss and compact design demonstrates the potential of this PCF sensor for biochemical sensing applications. These findings indicate its effectiveness for precise and reliable detection of sucrose concentration in real-world environments.
