Background: The rapid urbanization and aging of global civil infrastructure pose significant challenges to maintenance and safety management. Artificial intelligence (AI) has emerged as a transformative solution for enabling proactive, data-driven infrastructure management.
Objective: This study systematically evaluates AI-driven predictive modeling frameworks for smart civil infrastructure systems, identifying optimal techniques, performance benchmarks, and implementation strategies.
Methods: A comparative literature methodology was employed, analyzing peer-reviewed studies published between 2015 and 2024. Performance criteria including prediction accuracy, computational efficiency, maintenance cost reduction, and reliability indices were assessed across multiple AI model categories.
Results: Deep learning architectures—particularly Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs)—demonstrated superior prediction accuracy (93–95%) compared to classical machine learning models (84–89%). Integrated sensor-AI frameworks yielded an average maintenance cost reduction of 31.4%.
Conclusion: AI-driven predictive models represent a paradigm shift in infrastructure management. Future research should address data standardization, model explainability, and scalable deployment across heterogeneous infrastructure networks.