Abstract
—This study presents an intrusion detection system that utilizes strain-based optical fiber Bragg grating (FBG) sensors, that have been installed under the engineering bridge of the University of Johannesburg. The system uses different machine learning (ML) algorithms, including linear discriminant analysis (LDA), random forest algorithm, and neural networks specifically multilayer perceptron (MLP), to classify intrusions under varying conditions. The experimental setup is designed to monitor structural integrity and security and demonstrates the effectiveness of FBG sensors for intrusion detection. The comparative analysis of the ML classifiers reveals that the random forest algorithm achieves the highest accuracy at 90.89%, followed by LDA at 83.35%, and neural network MLP at 80.66%.These findings underscore the potential of combing FBG sensors with advanced ML algorithms for vigorous intrusion detection for critical infrastructure and environments.