A methodological approach to hybrid AI systems for real-time infrastructure monitoring in civil engineering

Abstract

Ensuring the safety and resilience of critical civil and transportation engineering infrastructure requires real-time, intelligent monitoring systems capable of detecting early signs of deterioration. Traditional Structural Health Monitoring (SHM) methods—primarily reliant on manual inspections or threshold-based sensor alerts—struggle to deliver the responsiveness, adaptability, and scalability demanded by modern urban environments in the fields of civil and transportation engineering. This paper introduces a hybrid Artificial Intelligence (AI) framework that integrates machine learning (ML), deep learning (DL), and rule-based reasoning within an edge–cloud architecture for real-time infrastructure monitoring. The system architecture consists of edge-level ML models, including Support Vector Machines and Random Forests, for fast anomaly detection; cloud-level CNN-LSTM networks for temporal pattern recognition; and a rule-based expert system to ensure interpretability and domain consistency across civil and transportation engineering use cases. Data from distributed IoT sensors is pre-processed, normalized, and fused using wavelet transformation, PCA, and statistical extraction methods. Metaheuristic optimization algorithms—Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Grey Wolf Optimizer (GWO)—are employed to fine-tune hyperparameters and select relevant features. Experimental results demonstrate high classification accuracy (up to 96.2%) at the edge, low prediction error (RMSE = 0.085) in cloud-based forecasting, and generalizability under optimization. The proposed hybrid AI system outperforms conventional SHM systems in speed, accuracy, and domain robustness, and is validated for real-world applications in civil and transportation engineering infrastructure.