A Hybrid Data-Model and AI-Driven Approach for Structural Monitoring in Hazardous Construction

This open access book addresses a critical challenge in modern construction: ensuring the safety of hazardous and complex engineering structures, such as super-tall buildings and large-span structures characterized by their slenderness and scale. The widespread use of these critical structures necessitates advanced safety monitoring and early warning systems. Traditional data-driven methods often fall short in meeting the demands for real-time, accurate, and proactive alerts under complex construction environments and extreme conditions. Therefore, research into hybrid data-model driven monitoring and early-warning technologies holds significant engineering importance.

(1) Hybrid Data-Model Driven Theory: A foundational framework is established, analyzing core models like Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory networks (BiLSTM), and AdaBoost. A novel CNN-BiLSTM-AdaBoost hybrid prediction model is proposed, along with an overall implementation framework.

(2) Hybrid-Driven Prediction for Tower Crane Response under Typhoons: A hybrid method is developed to predict tower crane displacement under extreme typhoons. An IoT-based monitoring system collects real-world data, while a Finite Element Method (FEM) model supplements extreme-scenario data. Predictions using pure data-driven and hybrid methods are compared.

(3) Real-Time Displacement Monitoring for High-Formwork Using Computer Vision: The M-DAVIM vision-based method is investigated. Controlled experiments quantify the impact of factors like light intensity, fog, camera angle, and vibration on measurement accuracy. Deployed at a real construction site in Ningbo, the system achieved sub-millimeter accuracy under optimal conditions (illuminance: 200-400 lux, target size >18 pixels) and demonstrated strong robustness, enabling real-time tracking of key nodal displacements.

(4) Hybrid-Driven Warning Threshold Update & Short-Term Response Prediction for High-Formwork: A three-module framework is proposed: a vision system for monitoring, a hybrid module for determining and dynamically updating safety warning thresholds, and a prediction module using the CNN-BiLSTM-Adaboost algorithm for one-hour-ahead displacement forecasting and construction load inversion.

Qiang Li, Ph.D., is an associate professor in the School of Civil Engineering at NingboTech University. He also holds administrative roles as Deputy Director of the Academic Affairs Office and the Center for Faculty Development, and Deputy Director of the Institute for Coastal Engineering Structures and Materials. He earned his Ph.D. in Structural Engineering from Zhejiang University in 2018. His primary research interests include low-altitude wind field safety, structural wind engineering, structural health monitoring and vibration control, digital twin technology, and smart construction and maintenance. Dr. Li has secured and led over ten significant research projects, including grants from the National Natural Science Foundation of China and key R&D programs in Ningbo, with total secured funding exceeding 3 million RMB. He has authored more than 30 SCI/EI-indexed journal articles and holds 20 authorized invention patents.