DIGITAL TWIN FOR STRUCTURAL HEALTH MONITORING OF CULTURAL HERITAGE: THE BUILDCHAIN DEMO-PILOT, PALAZZO PONIATOWSKI-GUADAGNI IN FLORENCE

Preserving the historical and cultural value of the built environment complying with the recent EU directives addressing resilience, sustainability, and energy efficiency in the building sector, is a modern challenge for Europe, as historical urban centres are prone to earthquakes and climate extremes events. Masonry buildings are particularly relevant in this context and the proper definition of mechanical parameters can be a crucial issue since they can vary in a wide range significantly affecting the outcomes of the assessment and the intervention strategies. A robust treatment of material uncertainties would require a substantial number of material tests, which can be both costly and time-consuming, as well as not aligned with the requirements of preservation. For this reason, it is extremely useful to rely on Bayesian inverse methods for the calibration of structural models based on limited measurements of the structural response. In the paper, the main steps towards the definition of digital twin for structural purposed of a heritage building are presented with reference to a significant case study, the Palazzo Poniatowski-Guadagni in Florence, a 18th century masonry building serving today as the headquarter of the local police. First, the development of Building Information Model (BIM) based on laser scanner survey data and the subsequent derivation of Finite Element (FE) model is shown. Then, a generalized Polynomial Chaos expansion (gPCE) surrogate model is introduced to reproduce the natural frequencies of the building. The surrogate model significantly decreases the computational time of the physics-based FE model, allowing the propagation of uncertainties in material properties, such as the elastic and Poisson’s ratio of masonry, relevant for the building’s dynamic response. Global sensitivity analyses are carried out using Sobol’ indices to assess the impact of input variability on the eigenfrequencies, evaluating the need for experimental campaigns and the set-up of a structural health monitoring (SHM) system. Replacing the deterministic FE solver with a gPCE surrogate model will facilitate quasi-real-time SHM. Therefore, the outcome will be a digital twin that serves as a tool for early warning and damage detection, relying on the Bayesian inference approach. The developed digital twin will be part of the BUILDCHAIN system, including Digital Building Logbook (DBL) and BIM, contributing to demonstrate the benefits of implementing innovative DBLs for management and preservation of cultural heritage.