Real Time Edge Intelligent IoT Framework for Uncertainty Aware Fiber Bragg Grating Sensing Using Hybrid Machine Learning Bayesian Compensation

Smart infrastructure, industrial IoT, cyber‐physical systems, smart and intelligent cities demand an accurate, real‐time, and reliable sensing capability for proper monitoring and decision‐making. Fiber Bragg Grating (FBG) sensing systems are used frequently in such an environment; however, the FBG sensing systems' accuracy is influenced by multi‐variable errors caused by strain‐temperature coupling, environmental disturbances, and drift. This paper presents a real‐time edge‐intelligent Internet‐of‐Things (IoT) framework for uncertainty‐aware compensation of FBG wavelength measurements using a hybrid machine– learning Bayesian approach. The system integrates an FBG sensor (∼1550 nm), a broadband SLED (> 40 nm bandwidth, ∼10 mW), and a high‐resolution interrogator (1510–1595 nm, < 0.5 pm resolution, up to 6 kHz), coupled with an edge computing unit for real‐time processing and internet connectivity. Wavelength‐domain data are processed locally at the edge, where machine‐learning models nonlinear errors and Bayesian inference provides uncertainty‐aware refinement. The compensated data are transmitted using lightweight protocols, enabling scalable and bandwidth‐efficient operation. Experimental results under coupled strain (0–2000 με) and temperature (25°C–80°C) conditions show RMSE reduction from 0.0127 μm to 0.0073 μm (42.6%), with peak error reduction of 58.3% and ∼35% drift reduction. The system achieves 3–5 ms processing latency, < 10 ms edge delay, and 25–45 ms end‐to‐end latency, along with 8–12× bandwidth reduction. These results demonstrate accurate, low‐latency, and scalable IoT‐based sensing with improved reliability and uncertainty awareness.