A medieval tower in Bologna is at risk of collapse, but AI and supercomputing may hold the key to its survival. Mark Venables explores how a high-fidelity digital twin, powered by CINECA’s Leonardo supercomputer and NVIDIA Omniverse, is transforming heritage conservation.
The medieval skyline of Bologna is defined by its towers, a legacy of an era when the city’s wealthy families built these structures as both symbols of power and defensive strongholds. The Garisenda Tower is among the most recognisable, standing alongside the taller Asinelli Tower. However, its future is uncertain. In 2023, unusual movements were detected, prompting urgent restoration efforts to prevent a potential collapse. An ambitious digital twin project is now at the centre of the preservation strategy, harnessing AI, supercomputing, and real-time data to safeguard the historic structure.
The role of digital twins in conservation
A digital twin is more than a static digital model. It is a dynamic, data-driven replica of a physical structure, continuously updated with real-world information to predict and analyse its behaviour. In the case of the Garisenda Tower, this means monitoring structural changes, assessing risks, and simulating possible interventions before they are applied in the real world.
“Digital twins are not just technological models; they are a fusion of data, technology, AI, and simulation capabilities,” Chiara Dellacasa, project manager in the supercomputing department at CINECA, explains. “They allow us to understand an asset’s structural behaviour and simulate outcomes in response to environmental changes or interventions. The key is ensuring that the digital model remains physically accurate, with continuous data flow between the physical and digital counterparts.”
The Garisenda Tower’s digital twin integrates data from over 100 sensors installed throughout the structure, measuring temperature, inclination, and deformation. While structural monitoring systems have existed for some time, the real innovation lies in how this data is being processed. High-performance computing and AI-driven analysis enable predictive modelling, helping authorities make informed restoration and risk mitigation decisions.
The digital twin does not merely track changes; it enables decision-makers to test restoration strategies virtually before applying them. AI-driven simulations evaluate different reinforcement techniques, predicting their effectiveness under various conditions, including extreme weather and seismic activity. This capacity to model interventions in a risk-free digital environment ensures that the most effective solutions are prioritised before physical work begins.
“Additionally, the digital twin serves as a historical archive, preserving vital structural information that can be accessed decades into the future,” Dellacasa adds. “Unlike conventional documentation, which often lacks real-time data integration, this approach provides a continually evolving record of the tower’s behaviour, allowing future conservationists to understand the impact of past interventions and environmental changes.”
Supercomputing as the foundation
Bologna is home to one of Europe’s most advanced supercomputing centres, CINECA, which provides the technological backbone for the project. At its core is Leonardo, the ninth most powerful supercomputer in the world, equipped with 14,000 GPUs for parallel computing.
“Leonardo’s capabilities allow us to run highly complex calculations at an unprecedented scale,” Dellacasa explains. “For instance, it can train a model like ChatGPT-3.5 in just one day. This processing power is essential for AI-driven physics simulations, data analysis, and real-time structural assessments.”
Leonardo’s infrastructure includes 500 computational nodes connected via high-speed optical fibre, ensuring rapid data transfer and processing. This allows the digital twin to simulate various scenarios, from minor environmental changes to catastrophic failure events, enabling pre-emptive action before structural integrity is compromised.
The computing power available at CINECA enables near-real-time processing of sensor data. If a shift in inclination is detected beyond a predetermined threshold, engineers can receive instant notifications, allowing them to analyse the cause and take immediate corrective action. This capability ensures the tower remains stable while restoration work progresses.
“The ability to process massive datasets also allows for greater simulation granularity,” Dellacasa continues. “Detailed 3D models incorporating high-resolution laser scans and geospatial data provide hyper-accurate visualisations of the structure’s evolution. By overlaying different datasets, conservationists can observe patterns that would be imperceptible in a single-dimensional analysis, such as micro-fractures spreading over time.”
Integrating AI and real-time data
One of the primary challenges in building the digital twin is data management. With vast amounts of real-time sensor data being collected, ensuring a single, unified source of truth is crucial. This is achieved through advanced AI models that process and interpret incoming information, identifying trends and anomalies that might otherwise go unnoticed.
“AI plays a critical role in enhancing the reliability of the digital twin,” Dellacasa explains. “Machine learning algorithms help us detect global correlations between factors affecting the tower’s stability. For example, we have found that temperature fluctuations influence the tower’s natural frequencies, highlighting the importance of thermal effects on structural dynamics.”
Predictive analytics are also being deployed to anticipate structural changes over time. Statistical forecasting models, such as ARIMA (Autoregressive Integrated Moving Average) and SARIMA (Seasonal ARIMA), detect trends in tilting patterns and predict potential risks. These insights enable proactive maintenance strategies rather than reactive interventions, ultimately extending the tower’s lifespan.
Recent advancements in AI have further improved the precision of these predictive models. “Deep learning techniques are now being applied to analyse historical data, cross-referencing past tilting behaviour with environmental conditions,” explains says. “This helps to identify patterns that might indicate a high-risk period for further destabilisation, allowing preventive measures to be implemented well in advance.
“AI is also enhancing material analysis by using physics-informed machine learning models. These algorithms can predict how different restoration materials interact with the tower’s existing brickwork, assessing durability and compatibility before any physical modifications occur. This minimises risk and ensures that preservation methods are tailored to the tower’s specific structural requirements.”
Bringing historical preservation into the digital age
The digital twin initiative is a collaborative effort between the Municipality of Bologna, the University of Bologna, and CINECA, with NVIDIA providing the technological framework. NVIDIA Omniverse is the core data integration and visualisation platform, using Universal Scene Description (USD) to create an interactive and photorealistic tower representation.
“The power of Omniverse lies in its ability to consolidate diverse datasets into a single, interactive environment,” Dellacasa says. “We can use AI-driven tools to analyse historical sites, simulate environmental effects such as weathering or seismic activity, and create immersive virtual reality experiences for education and museum applications.”
The project also leverages high-resolution laser scanning, photogrammetry, and structural modelling techniques. Conservationists can assess specific vulnerabilities in detail by segmenting architectural elements, such as walls, columns, and sculptures, into individual components. High-resolution digital meshes track changes over time, identifying cracks, deformation, and erosion patterns that might indicate structural weakening.
Omniverse enables real-time collaboration between experts across multiple disciplines. Engineers, architects, and AI specialists can work within the same virtual space, instantly updating models and sharing insights. This level of integration accelerates decision-making and ensures that all stakeholders have access to the latest, most accurate data.
The future of AI-driven heritage conservation
While the primary objective of the Garisenda Tower’s digital twin is to ensure its preservation, the technology has far-reaching implications beyond heritage conservation. The methodologies developed for this project could be applied to infrastructure monitoring, urban planning, and even climate modelling.
“Digital twins represent a significant shift in how we approach preservation,” Dellacasa concludes. “They enable us to move beyond reactive conservation methods towards a predictive, data-driven approach. This is not just about safeguarding historical sites but redefining how we monitor and manage structural integrity across a wide range of applications.”
As the project advances, integrating AI, high-performance computing, and real-time data will continue to play a pivotal role in preserving one of Bologna’s most iconic landmarks. The Garisenda Tower may have stood for nearly a millennium, but its future will be shaped by some of the most cutting-edge technologies of the modern era.
