Intelligent structural health monitoring system and method based on fiber optic bag sensor networks (FBG) and xls™-meanpool deep learning algorithm

Abstract

The useful solution refers to a structural health monitoring (SHM) system and methodology based on fiber optic sensor technology (FBG) integrated with an Internet of Things (IoT) platform. The system includes a broad-spectrum SLED light source (peak spectrum 1550 nm, power ≥12 mW) with an integrated optical circulation system, a temperature-compensated FBG sensor network, an optical signal decoder, and a data logger for real-time data transmission. The core of the solution is a damage diagnosis method that applies the xLSTM-MeanPool deep learning algorithm, implemented through the following steps: signal patching, extraction of hidden states using the xLSTM architecture, and synthesis of sequence-level features using masked mean pooling. The solution allows for the complete elimination of electromagnetic interference thanks to the FBG fiber optic sensor system, optimizes computational efficiency through segmented embedding techniques, and achieves structural condition diagnostic accuracy of 97.40% on FBG data and 97.37% on complex real-world Z24-9Setup data. This useful solution provides the ability to detect early signs of potential damage to transportation infrastructure in real time with significantly higher accuracy (up to 31.76% on real-world data) compared to current Transformer architectures.