A Multi-Scale Signal Fusion Framework for Intelligent Structural Stress Monitoring Using Distributed Optical Fiber Sensing

To more efficiently address the limitations of the distributed fiber optic sensing in structural stress monitoring, such as limited spatial resolution, severe coupling of multiple physical quantities, and insufficient multi-scale perturbation identification capabilities, this study proposes a multi-scale signal fusion intelligent structural stress monitoring framework based on distributed fiber optic sensing. This framework integrates Differential Pulse Pair (DPP) super-resolution demodulation, joint modeling of multiple physical quantities, multi-scale mode decomposition, and instantaneous phase feature extraction methods. This innovation enables the coordinated sensing and high-precision inversion of temperature, strain, stress, and micro-perturbation information. At the physical modeling level, this model constructs a unified multi-physical quantity coupling model based on Raman scattering temperature demodulation, Rayleigh coherent scattering vibration sensing, and Brillouin strain response. At the signal processing level, the model introduces a differential pulse pair equivalent compression mechanism and a threshold weight compensation model to suppress spatial aliasing errors. Furthermore, as an enhancement, instantaneous phase change rate features are constructed based on Hilbert transform to improve the ability to identify micro-amplitude high-frequency disturbances and local damage. Finally, multi-scale mode decomposition and feature fusion are used to achieve collaborative modeling and robust demodulation of multi-scale structural responses. On a multi-physics experimental platform, the proposed method was systematically validated under static load, dynamic disturbance, thermo-mechanical coupling, and local damage evolution conditions. Experimental results show that even with a hotspot scale only$16 \%(0.08 \mathrm{m})$of the system's theoretical spatial resolution, the temperature peak recovery rate of the proposed method still exceeds 88 %. Under high load conditions (20-30 kN), the stress inversion error is stably controlled within 2 MPa, while the maximum error of the comparative method exceeds 5 MPa. Under low-amplitude high-frequency vibration conditions of 30-50 Hz, the detection error of the proposed method is reduced by approximately 37 % compared to the traditional amplitude domain method. Under temperature fluctuations of$\pm 20^{\circ} \mathrm{C}$, the stress inversion drift amplitude is compressed to within 1.3 MPa. In the stage where the microcrack width is less than 0.2 mm, the damage identification accuracy is improved by more than 20 percentage points compared to the comparative method.