International Journal of Scientific Research and Engineering Development

This paper presents a five-layer digital twin (DT) architecture for real-time bridge construction monitoring, integrating Internet of Things (IoT) sensor fusion, deep learning inference pipelines, and BIM–finite element analysis (FEA) coupling. The physical layer deploys fibre Bragg grating (FBG) strain sensors, MEMS accelerometers, RTK-GNSS receivers, and thermistor arrays at critical structural locations, transmitting data via MQTT over a hybrid LoRaWAN/5G edge-cloud topology. The data layer implements an Apache Kafka streaming pipeline with edge pre-processing on NVIDIA Jetson Orin modules, achieving sub-200 ms sensor-to-dashboard latency. The digital modelling layer couples an IFC 4.3–compliant BIM model with a Midas Civil / LUSAS finite element model through automated Bayesian model updating using Markov Chain Monte Carlo (MCMC) sampling, calibrating stiffness parameters in real time against measured sensor responses. The AI analytics layer deploys four deep learning models: (1) a stacked Bi-LSTM network for construction progress forecasting (MAE = 2.3% on benchmark data); (2) a variational autoencoder with multi-head self-attention (VAE-Transformer) for structural anomaly detection (F1 = 0.94); (3) a YOLOv8n object detection model for CDM 2015–compliant safety monitoring (mAP@0.5 = 0.91); and (4) an XGBoost ensemble for concrete compressive strength prediction from mix design and curing sensor data (R² = 0.96). All models are containerised in Docker and orchestrated via Kubernetes for scalable deployment. The architecture is designed for the United Kingdom’s infrastructure context, compliant with ISO 19650, the Gemini Principles, and the Design Manual for Roads and Bridges (DMRB). Expert validation (N = 12) yielded a mean technical feasibility score of 4.25/5.00. The system architecture, data schemas, model hyperparameters, and inference pipeline specifications are detailed to enable reproducibility.

Moustafa Metwally,"AI-Driven Digital Twin Architecture for Real-Time Bridge Construction Monitoring: IoT Sensor Fusion, Deep Learning Pipelines, and BIM-FEA Integration for UK Infrastructure" International Journal of Scientific Research and Engineering Development, V9(2): Page(253-258) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.