FOURIER aims to leverage the power of rapidly advancing AI technologies to transform the critical infrastructure (CI) industry. Our mission is to develop innovative AI-driven approaches that will improve inspection procedures, Structural Health Monitoring (SHM) data processing, and resilience decision-making across infrastructure systems.
To achieve this mission, the project is focused on several measurable Research Objectives:
These goals will drive the FOURIER project towards a future where infrastructure systems are more efficient, resilient, and adaptive, benefiting both the industry and society as a whole.
Hosting institution: Delft University of Technology (TU Delft)
Supervisors: M. Nogal (TU Delft), P. Clemente (ENEA)
Objectives: Develop ASI technologies for autonomous inspection and system monitoring using AI algorithms for remote control and automated data processing in advanced inspection systems.
Description: The development of AI has three stages: artificial narrow intelligence (ANI), artificial general intelligence (AGI), and artificial super intelligence (ASI). ANI is less intelligent than humans and requires intervention. AGI matches human intelligence and operates independently. ASI surpasses human intelligence, performing tasks beyond human capability. This PhD thesis aims to develop ASI-level computational models for damage detection, structural health assessment, and predictive analysis. ML methods will be proposed for model updating, diagnostics, and data interpretation in structural health monitoring (SHM) systems. AI-based SHM methods, like those from the University of Texas at Arlington, use sensors to assess bridge health. Modern infrastructures incorporate weight-in-motion systems with sensors measuring vibrations, strains, and deflections. By analyzing these responses, they estimate vehicle weights and their impact on structural integrity. ML techniques refine load parameters and provide a clearer structural assessment, helping prevent failures like the 2018 Ponte Morandi collapse. An IoT-based SHM system can also detect damage and monitor structural behavior.
Secondments: National Agency for New Technologies, Energy and Sustainable Economic Development-ENEA (IT), Veiligheidsregio Utrecht (NL)
Hosting institution: Politecnico di Torino (POLITO)
Supervisors: G. P. Cimellaro (POLITO), A. Cardoni (POLITO), D. Antunes (FACTOR SOC), D. Inaudi (SMARTEC), G. Tsionis (JRC)
Objectives: Integrate AR with advanced non-contact sensors to enhance inspections using Deep Learning models for multimodal analysis, tested in real-world applications.
Description: Current infrastructure inspections still rely on human site visits aided by low-cost RGB cameras, which lose 3D information when projecting to 2D. Studies show that non-conventional imaging sensors like depth, thermal, and hyperspectral cameras provide critical data that RGB cameras cannot capture, but the structural health monitoring (SHM) community has yet to fully utilize them. Research indicates that multi-modal DL models combining heterogeneous data improve detection accuracy. This study will use non-contact sensors such as LiDAR, LDV, and thermal cameras to develop real-time multi-modal DL models, enhancing data extraction over traditional vision-based techniques. The processed data will be visualized via AR wearables like Microsoft HoloLens 2, equipped with multimodal sensors and a minicomputer. A standalone prototype will be developed at POLITO, tested at FACTOR SOC, and demonstrated on real infrastructures using data from SMARTEC.
Secondments: Factor Social (PT), Smartec SA (CH)
Hosting institution: Politecnico di Torino (POLITO)
Supervisors: G. P. Cimellaro (POLITO), A. Cardoni (POLITO), G. Tsionis (JRC), E. De Iuliis (ASI_Europe), P. Furtner (VCE)
Objectives: Develop a methodology to: (i) use UAVs for rapid data collection on buildings and CIs, (ii) create digital twins and calibrate numerical models, and (iii) analyze progressive collapse risks to estimate residual resistance capacity.
Description: Many strategic buildings in Europe have exceeded their lifespan, posing a risk of progressive collapse, economic loss, and safety concerns. Risk analysis requires detailed computational models with large, accurate datasets, but current visual inspections by engineers are costly and time-consuming for resource-limited authorities. This PhD project aims to deploy a swarm of UAVs equipped with hybrid sensors (e.g., hyperspectral cameras, LiDAR) to collect data automatically. ML algorithms will process this data to identify buildings' geometrical, structural, and thermal parameters. A prototype system with three AI-driven drones will prevent collisions while collecting and processing data. The information will be stored in a database and used to calibrate computational models for progressive collapse analysis. Software training for this analysis will take place at ASI_Europe, with results integrated into a G.I.S. model for stakeholder access.
Secondments: Vienna Consulting Engineers (AT), Applied Science International Europe, srl (IT), Joint Research Center – Ispra (IT)
Hosting institution: Technical University of Munich (TUM)
Supervisors: D. Straub (TUM), M. Schubert (MATRISK), G. P. Cimellaro (POLITO)
Objectives: To develop AI-based explainable predictive maintenance (PM) solutions based on heuristic.
Description: Current infrastructure inspection follows time-based maintenance plans, but predictive maintenance (PM) can reduce costs and improve safety. AI-driven reinforcement learning (RL) has been used for PM planning, such as deep neural networks with an actor-critic architecture. While effective, deep RL solutions face challenges due to their black-box nature. This project aims to develop AI-based explainable PM solutions using heuristic approaches from the TUM group, integrating theory-guided ML for better prognostics. In year 1, heuristic maintenance strategies will be developed and compared to deep RL solutions through numerical experiments. Theory-guided ML algorithms will also be tested on benchmark prognostics datasets. In years 2 and 3, real-life data will be used to refine and integrate these algorithms into optimal heuristic maintenance planning.
Secondments: Matrisk (CH), Politecnico di Torino (IT)
Hosting institution: Albert Ludwig University of Freiburg (UFR)
Supervisors: S. Hiermaier (UFR), A. Reiterer (UFR), I. Häring (Fraunhofer), G. P. Cimellaro (POLITO)
Objectives: To overcome limitations related to low-resolution structural health monitoring (SHM) and inspection data developing AI-based techniques.
Description: The quality of monitoring data significantly affects the accuracy of AI-based defect detection algorithms. In visual inspections using drones and open-source spatial data (e.g., OpenStreetMap, satellite data), image and video quality is often limited. This research will use AI-based multi time-series data and video super-resolution techniques to reconstruct high-resolution images and videos from low-resolution inputs, maximizing information gain and minimizing uncertainty. This approach reduces the need for high-resolution cameras and sensors. The goal is to develop AI-based asset classification and information extraction for assessing mechanical health, identifying degradation, failures, and damage from malicious events (e.g., impacts, drone-delivered damage). The techniques will be developed for buildings and green infrastructure subsystems, validated using data from partners and municipalities in case studies.
Secondments: Eletrizitätswerke Schönau Enerige GmbH - EE Infratec GmbH (DE), Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institute, EMI (DE), Politecnico di Torino (IT)
Hosting institution: Infra Plan consulting d.o.o.
Supervisors: I. Stipanović (INFRAPLAN), M. Bačić (UNIZAG GF), M. Fragiadakis (NTUA)
Objectives: To overcome lack of data implementing procedures for generating large-scale synthetic data via digital twins and generative adversarial networks.
Description: AI-based monitoring of distributed CIs requires large training datasets for accurate predictions, but a lack of field data on damage events limits application. Manual data annotation is time-consuming, error-prone, and impacts model performance. This task addresses these challenges by using synthetic data generated from simulations of physical engineering models (digital twins), incorporating uncertainties, pre-stress (e.g., aging), external loading, failure response, and recovery. Generative adversarial networks (GANs) will produce large datasets from digital twin simulations, which will be used to train AI algorithms to assess multi-time series data from infrastructure elements. In the final step, these AI algorithms will be validated to identify physical degradation parameters from multi-time series sensor data.
Secondments: National Technical University of Athens (EL), Technical University of Munich (DE), University of Zagreb Faculty of Civil Engineering (HR)
Hosting institution: National Technical University of Athens (NTUA)
Supervisors: M. Fragiadakis (NTUA), V. Tsoukala (NTUA), M. Nogal (TU Delft), I. Stipanović (Infraplan)
Objectives: Implement AI techniques that enable inverse problem solving and automatic DT updating.
Description: AI algorithms, like neural networks, are highly effective in inverse problems, where the goal is to infer the causes of observed outcomes by deducing complex relationships between input and output variables. These methods help reconstruct underlying parameters or conditions from observed data. AI also plays a key role in updating digital twins (DTs), which are virtual models of physical assets that need continuous refinement to remain accurate. In this task, reinforcement learning and deep learning algorithms will be developed to automatically update DTs by integrating new data and adjusting the virtual model to reflect changes in the physical entity. This combination of AI and inverse problem-solving, along with dynamic DT updating, enables more efficient and accurate decision-making in various sectors, including healthcare and engineering. The developed algorithms will be tested and validated on different CIs, such as bridges, harbor structures, and chemical industry facilities.
Secondments: Delft University of Technology (NL), Infra Plan consulting d.o.o. (HR)
Hosting institution: Technical University of Munich (TUM)
Supervisors: D. Straub (TUM), M. Nogal (TU Delft)
Objectives: Integrate monitoring data and probabilistic digital twins for resilience management of communities.
Description: Spatial data and information are crucial for managing the risk and resilience of urban areas. Continuous monitoring data from various sources can be used to update information and create real-time representations, supporting optimized decision-making for tasks like traffic management and emergency response. However, current urban information systems often do not account for uncertainty, especially in emergencies. This project will explore the use of a probabilistic digital twin framework, similar to those in the aerospace sector, and adapt it for urban systems. The focus will be on emergency response to natural or man-made hazards, integrating data from satellites, urban monitoring systems, and networks into a digital representation using a Bayesian framework.
Secondments: Matrisk (CH), Delft University of Technology (NL)
Hosting institution: VCE Vienna Consulting Engineers ZT GmbH (AT)
Supervisors: P. Furtner (VCE), M. Nogal (TU Delft), D. Inaudi (Smartec)
Objectives: Integrate statistical, big data strategies into an structural health monitoring (SHM) framework for damage detection and prediction for long-term monitoring.
Description: Due to the complexity and variability of SHM, large amounts of data are produced, making big data strategies essential. This task will integrate statistical big data strategies with SHM systems to detect and predict structural damage in harsh, variable environments over long monitoring periods. The big data SHM framework will utilize techniques like dynamic time warping, singular value decomposition, factor analysis, and maximum likelihood statistics. It will be tested through short-term (hours to days) and long-term (multi-year) campaigns on infrastructures exposed to various damage and environmental conditions. The project will use data from different sensors, including fiber optics. Developed tools will be applied and tested at SMARTEC using real-world CI data and demonstration case studies in Utrecht and Arrábida Territory.
Secondments: Delft University of Technology (NL), Smartec SA (CH)
Hosting institution: Factor Social (FACTOR SOC)
Supervisors: J. Palma-Oliveira (FP-UL), D. Antunes (FACTOR SOC), M. Rhoen (VRU)
Objectives: Develop tools to monitor and evaluate the social infrastructure resilience for improved decision-making.
Description: Community resilience relies on both social and physical assets. This task aims to (a) identify the social dimensions of resilience, such as risk perception, social capital, preparedness behavior, decision-making, crisis communication, and public behavior during crises, and (b) develop and test methodologies to monitor these aspects under various scenarios considering uncertainties and stress levels. A decision support tool will be created to: (i) incorporate monitoring data to assess the resilience of social infrastructure, focusing on the psychosocial and behavioral perspectives of users and stakeholders, and (ii) analyze data to plan improvements. The tool's inputs will come from consultations with communities, municipal representatives, and experts, as well as from case study data in Utrecht and the Arrábida Territory.
Secondments: Faculty of Psychology of the University of Lisbon (PT), Veiligheidsregio Utrecht (NL), Energy and Environment Agency of Arrábida (PT)
Hosting institution: Delft University of Technology (TU Delft)
Supervisors: M. Nogal (TU Delft), D. Straub (TUM)
Objectives: Develop a multi-fidelity digital twin (MFDT) of interdependent CIs and develop new AI algorithms to determine the MFDT level of fidelity.
Description: Managing infrastructure is challenging due to increasing maintenance costs and the uncertainty about their current state. Detecting deterioration early can help implement effective decision-making strategies. This task focuses on developing an AI surrogate modeling procedure that condenses (structural health monitoring) SHM data into a probabilistic, time- and state-continuous surrogate model, updated with real-time monitoring data. The model will account for uncertainties from conflicting, vague, or incorrect data, and propagate them accordingly. Structural numerical analysis results will be integrated into a resilience framework, including recovery models and cost functions. This will allow decision-makers to make cost-efficient life cycle decisions in maintenance, repair planning, and design processes, considering component lifetimes and financial constraints.
Secondments: Technical University of Munich (DE), Veiligheidsregio Utrecht (NL)
Hosting institution: Albert Ludwig University of Freiburg (UFR)
Supervisors: S. Hiermaier (UFR), P. Furtner (VCE)
Objectives: Improve understanding and resilience quantification of CIs through causal inference analysis.
Description: Critical infrastructure systems are often analyzed through observations, but the effect of interventions and improvements is less explored. Understanding the system’s causal relationships is essential for determining these effects. This work focuses on the application of smart metering devices for smart homes and industry prosumers. First, causal diagrams are built from raw data sources, aggregating health status, policies, and simulation data. Then, causal inference is conducted through quantitative causal diagrams. The project will answer key questions: (i) what monitoring signals reveal about the system's health, (ii) how the system behaves with component degradation or improvement, and (iii) how interventions contribute to system improvements, considering other ongoing processes. This three-step causal ladder approach will identify accurate causal relationships and quantify uncertainties. The results will guide improvements in smart metering devices, focusing on cost-efficient component upgrades that enhance reliability, maintainability, and subsystem performance.
Secondments: Eletrizitätswerke Schönau Enerige GmbH - EE Infratec GmbH (DE), VCE Vienna Consulting Engineers ZT GmbH (AT)
Hosting institution: Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institute, EMI (Fraunhofer)
Supervisors: I. Häring (Fraunhofer), A. Stolz (UFR), D. Straub (TUM), S. Hiermaier (UFR)
Objectives: Use adaptive HMM approaches to reproduce and predict the dynamic behavior of power network prosumers and achieve efficient and sustainable use of storage and smart metering devices.
Description: Power grid infrastructures are constantly evolving due to policies, regulations, and the growing capabilities of prosumers. A key challenge is how to model these actors, especially without their cooperation. This research focuses on models that incorporate novel storage devices used by prosumers, such as redox flow batteries and heat storage systems. The project will explore adaptive hidden Markov models (HMMs) to predict and understand the dynamic behavior of prosumers with these storage technologies. Input data will include power consumption, generation time-series, and costs. The goal is to reduce hidden variables in the HMM using a minimal set of explanatory parameters, such as strategy hierarchy, storage characteristics, and presumption patterns. The research aims are to modify monetary incentives for prosumers to optimize the use of storage devices, and to integrate HMMs into grid simulations for efficient adaptation. The project also seeks to support smart metering providers by enhancing operation and maintenance of devices using HMMs, which can operate offline and aid in smart meter monitoring and maintenance.
Secondments: Eletrizitätswerke Schönau Enerige GmbH - EE Infratec GmbH (DE), Technical University of Munich, TUM (DE), University of Freiburg, UFR (DE)
Hosting institution: Factor Social (FACTOR SOC)
Supervisors: J. Palma-Oliveira (FP-UL), I. Linkov (FACTOR SOC), G. P. Cimellaro (POLITO)
Objectives: Development of a framework to evaluate the social and market acceptability of Digital Inspection Tools.
Description: Tools developed by academic organizations often fail to meet stakeholder needs and end up unused in a library. The PhD project aims to create a framework for evaluating the social and market acceptability of digital inspection tools. The research will involve systematizing literature on technology acceptance and conducting interviews with stakeholders, including infrastructure managers and inspectors. These efforts will lead to the development of a comprehensive model for assessing the social and market acceptance of digital technologies for infrastructure inspection. This model will be applied to inspection tools being developed in three parallel PhD projects. A secondment at FP-UL will provide expertise in social science research methods, while Factor Soc will offer access to various stakeholders in Portugal. Additionally, a secondment at ENEA will offer insights into technology transfer to the market.
Secondments: Faculty of Psychology of the University of Lisbon (PT), Politecnico di Torino (IT), Energy and Sustainable Economic Development-ENEA (IT)
Hosting institution: Infra Plan consulting d.o.o. (INFRAPLAN)
Supervisors: I. Stipanović (INFRAPLAN), M. Bačić (UNIZAG GF)
Objectives: Improve decision-making in the management of transportation networks using probabilistic vulnerability assessment supported by AI.
Description: The impacts of climate change on transport infrastructure networks have become more frequent and severe, as shown by flooding, landslides, and road failures in Croatia and Slovenia in August 2023. These issues are often linked to substructure instability caused by extreme precipitation and flooding. The PhD project will focus on assessing the vulnerability of geotechnical structures (such as foundations, embankments, and support structures) within transport networks. It will integrate various levels of monitoring, from satellite and UAVs to embedded sensors, into probabilistic assessments. In-situ and numerical modelling data will inform the development of AI models, which will continuously improve with new data. Vulnerability assessments and consequence analyses will contribute to a decision optimization framework for risk management. This framework will be validated through case studies in rural, urban, and highway settings, incorporating socio-economic context and transportation network resilience.
Secondments: University of Zagreb Faculty of Civil Engineering (HR), InGEO BV (NL)
Under the Grant Agreement no. 101169429