Structural engineering

Integrating physics, data and machine learning for structural assessment and prediction under uncertainty

Ageing, damage and climate change challenge the safety and sustainability of infrastructure. Cost-effective repair, maintenance and life-cycle management require reliable assessments, yet real-world structural behaviour is complex, monitoring data are often limited or noisy, and physical models are uncertain.

His research integrates physics-based modelling, data and machine learning within a unified framework that accounts for these uncertainties, supporting structural identification, damage assessment, prediction and informed maintenance decisions.

It aims to develop efficient physics-informed neural networks and neural operators for structural identification, health monitoring and prediction, with quantified uncertainties.

Key questions concern learning from sparse data, accelerating inference through reduced-order modelling, and providing reliable uncertainty estimates for decision-making. Future goals include multiscale and multiphysics applications and experimentally validated real-time model updating and prediction.

Ruy Chang combines field and laboratory measurements with finite element and reduced-order models, physics-informed neural networks and neural operators. Bayesian inference supports model updating and uncertainty quantification in parameters and predictions.

The methods are evaluated using numerical benchmarks and experimental and monitoring data. Building on his research at Tongji University, Penn State and ETH Zürich, this work connects structural engineering, computational mechanics and scientific machine learning, with future collaborations extending validation and applications across disciplines.

Potential applications include structural assessment, damage identification, vibration mitigation and monitoring systems that update predictions as new data become available. Future extensions include multiscale modelling, linking material damage to structural response, and multiphysics modelling, such as coupled wind–wave–structure interaction in offshore wind turbines.

Expected impacts include faster assessment, better-informed maintenance decisions and more reliable life-cycle management. These advances could extend infrastructure service life, reduce costs, material consumption and carbon emissions, and improve safety and resilience while supporting the digital and energy transitions.
 

The architectural potential of tensegrity structures is proven. Yet, paradoxically, very few real construction projects have been built around the world. The main reasons are complex construction processes, lack of design and optimization guidelines, and excessive self-weight due to the pre-stress needed to guarantee stiffness and dynamic behavior. Our research in this field is aimed at removing the barriers that still prevent the use of tensegrity in the construction sector. With the aim of defining reliable design, calculation and construction methods, we carry out numerical developments and compare them with test results on full-scale prototypes.

Our research focuses on advancing structural engineering through large-scale experimental testing, aimed at better understanding the real-world behaviour of civil infrastructure. These tests are designed to bridge the gap between theoretical models and actual structural performance, enabling the validation of novel materials, systems, and design approaches. Full-scale or near full-scale specimens are subjected to complex load combinations, allowing us to observe failure mechanisms, time-dependent responses, and structural resilience under realistic conditions.
Recent and ongoing investigations include concrete elements reinforced with thermoplastic GFRP reinforcement, shape-memory alloy rebars, or hybrid systems, as well as low-carbon materials. These experimental campaigns provide high-fidelity local-level (strain) data that feed into computational model calibration, support the development of design codes, and inform resilient infrastructure strategies. The work is carried out in collaboration with academic and industrial partners, contributing to safer, more sustainable construction practices.
We often launch international blind prediction competitions associated to some of the large-scale experimental tests, as you can see here or here.
 

Humanity faces major environmental challenges, and using a maximum of timber in the construction industry is part of the solution. Building with timber, but also planting trees, re-create our past forests and extending the remaining ones is a priority. Our research in timber engineering is driven by the conviction that the use of minimally processed wood in the construction sector is likely to make it more competitive compared to concrete and steel. Our researches therefore cover a number of themes related to the use of logs for buildings : non-destructive mechanical characterization of logs, steel free connections, timber slabs, etc.

Our research in nonlinear and dynamic structural modelling addresses the complex behaviour of civil structures under extreme or time-dependent loading conditions. By accounting for material nonlinearity, geometric effects, and dynamic interactions, we aim to provide realistic simulations of how buildings, bridges, and infrastructure perform under seismic events, impact, fatigue, and long-term degradation. Particular focus is placed on mechanics-based models and numerical models that better capture the local structural response, capturing failure mechanisms such as cracking, crushing, buckling, or bond-slip phenomena. As an example, we have recently developed HybriDFEM, a new hybrid discrete-finite formulation enabling coupling with classical FEM.
These high-fidelity models are validated through experimental data and are used to support performance-based design and risk-informed assessment. Our ongoing work leverages both traditional numerical methods and data-enhanced approaches, including model updating, inverse identification, and hybrid physics–AI frameworks. This research contributes to developing more accurate predictive tools, improving the resilience and sustainability of critical infrastructure in both new design and retrofitting scenarios.
 

Reciprocal structures are of indisputable architectural interest, and make it possible to build structures with considerable spans using small components, such as logs. Our research is aimed at establishing the stability, design and calculation rules specific to these structures, as well as designing assemblies that can be reliably modeled. Our research is both numerical and experimental, with tests on large-scale prototypes.

To address the growing demands for sustainability, durability, and performance in civil infrastructure, our research explores innovative reinforcement materials as alternatives to, or complements for, conventional steel reinforcement in concrete structures. We have been focusing on fibre-reinforced polymer (FRP) bars and shape memory alloys, each offering unique benefits such as corrosion resistance, light weight, tailored mechanical properties, and reduced environmental impact.
These materials are studied both experimentally and through advanced modelling to assess their structural behaviour, durability, bond characteristics, and long-term performance under realistic service conditions. Particular attention is given to hybrid reinforcement systems and low-carbon concrete combinations to develop integrated solutions for the next generation of resilient and sustainable concrete infrastructure. Our work contributes to updating design guidelines and supports the safe implementation of these materials in practice.
 

Agrivoltaic structures can take various forms, depending on the type of crop envisaged: high structures with long spans or low structures such as ground frames or vertical barriers. Our researches focus on developing different types of optimal structures, cable-stayed or not, with an emphasis on the use of wood materials, and with various constraints such as minimizing the size of foundations to reduce intervention in cultivable soils. We are also working on optimizing structures for non-agricultural photovoltaic power plants.

Using high payload drones for the construction of buildings?

The aim of this research was to lay the necessary groundwork, still not explored elsewhere, in order to prove the feasibility of building real-scale structures, in particular masonry structures, with high payload drones.

Droxels…the universal construction component?

We created the droxels while searching for a structural component able to provide four essential features required by robot-based construction:
•    They can faithfully reproduce any 3D structural form with a good structural stability;
•    They have a large laying tolerance for easy interlocking;
•    They avoid the need for scaffolding in most cases;
•    Their size can be scaled up or down to reach the desired surface aspect.
Droxels were initially developed to build with flying robots, but they also have the potential to cover a large range of “do-it-yourself / home-handyman” applications, including garden/farm/industrial shelters, temporary/emergency structures, artistic structures, tiny houses, small footbridges and retaining walls.