Enhancing Fiber Bragg Grating Sensor Signal Prediction via Edge Computing in Sensor Networks
Lin Liu, Huang Li-Yuan · 2024
This paper endeavors to advance the prediction of Fiber Bragg Grating (FBG) sensor signals using data-driven edge computing, thereby extending the utility of data exchange-based edge computing within sensor networks. The initial focus entails devising a sensor network system founded on data exchange principles seamlessly applied to a highway construction project. Subsequently, an algorithm is introduced to predict anomaly signals from FBG sensors, mitigating errors stemming from external environmental interferences affecting the sensors. Temperature data from the sensors is then leveraged for multi-depth temperature detection (2 cm, 10 cm, and 20 cm), revealing anomalous temperature readings of 4.8°C, 1.5°C, and -0.25°C. Results attest to the efficacy of the proposed data exchange-driven edge computing method in FBG sensor signal prediction, demonstrating a true positive rate of 90.22% and a false negative rate of 9.0%. Following data training, the prediction algorithm yields a coefficient of determination of 0.876, emblematic of the successful integration of data exchange-based edge computing into FBG sensor signal prediction. Remarkably, edge computing markedly reduces data transfer latency and alleviates network load, surpassing traditional cloud-based data transfer in processing. Moreover, edge computing showcases exceptional real-time capabilities, facilitating more timely and precise FBG sensor result prediction.