SMM at EMI 2025: Pioneering Resilience in Cyber-Physical Systems
Researchers from the Chair of Structural Mechanics and Monitoring (SMM) and the Future Resilient Systems program of the SEC presented four contributions at the 2025 Engineering Mechanics Institute (EMI) Conference. Their work advances the resilience of cyber-physical systems through innovations in mobile sensing, interpretable AI, and digital twin technologies—demonstrating the group’s interdisciplinary leadership at the intersection of engineering and data science.
The 2025 EMI Conference, organized by the American Society of Civil Engineers, convened international scholars to share the latest developments in engineering mechanics. SMM researchers contributed four presentations that highlight our ongoing efforts within the Future Resilient Systems (FRS) programme to develop intelligent tools for monitoring, assessment, and decision support in civil infrastructure:
Robotic Sensing for Bridge Health Assessment
Xudong Jian presented a mobile sensing approach for bridge monitoring, using a low-cost robot equipped with vibration sensors. The system was validated across 12 bridge spans in collaboration with Singapore’s Land Transport Authority (LTA), offering a scalable and cost-effective solution for structural health diagnostics. Jian also co-chaired the EMI mini-symposium on Structural Identification and Damage Detection on behalf of Prof. Eleni Chatzi.
Multi-Resolution Neural Networks for Structural Identification
Kiran Bacsa showcased a deep learning approach based on multi-resolution convolutional neural networks to extract structural and vibrational features from measured data. The method operates in both supervised and unsupervised settings, harnessing spectral biases in neural networks to enable interpretable, data-driven structural monitoring.
Graph Neural Networks for Digital Twins of Electrical Networks
Huangbin Liang presented a physics-encoded graph neural network (PeGNN) developed for real-time digital twins of interconnected electrical components. Validated through full-scale seismic tests, the model captures dynamic interactions critical for post-earthquake assessment and decision-making in complex infrastructures.
Physics-Guided Neural ODEs for Scientific Discovery
Wei Liu introduced Structured State-Space Neural Ordinary Differential Equations (S3NODEs), a framework that embeds physical structure into neural architectures. The approach improves prediction accuracy and interpretability, supporting applications in scientific modeling and engineering diagnostics. Results were validated on synthetic and real-world datasets.
