3D Semantic Digital Twins: Data Streams and Ontologies for Risk Prediction in Heritage Contexts

Abstract. Sensor data are core for real-time monitoring in heritage contexts, allowing to ground the synchronization between the real asset and its Digital Twin (DT). Such synchronization can then be leveraged to perform risk prediction with a time aware reasoning process, which can be crucial for heritage prevention. Nonetheless, sensor data by itself lack both geometrical and semantic information which can also be pivotal in spotting such incipient degradations. For this reason, mature DT approaches need to tie these three main sources of knowledge and to conduct temporal reasoning in a form which can consider them altogether. This paper extends the 3DOnt Framework, originally designed to link geometries and semantics, by integrating sensor data streams and enabling time-aware reasoning over live updates, thus turning it into a Semantic-Geometric Digital Twins framework. The extension includes an expansion of the 3DOnt backbone ontology to manage sensor data alongside raster-encoded and other static information, to model sensors as individuals within the 3D Graph and to link them to the macro-objects they monitor. Furthermore, 3DOnt’s reasoning processes are enhanced to exploit dynamic sensor inputs, incorporating the temporal persistence of risk factors in prediction tasks. This extension is evaluated on the 3D Graph of a portion of Centro Culturale Santa Chiara, in Trento (Italy). With historical records simulating dynamic sensor data streams concerning temperature and humidity, we integrate this numerical information with semantic-geometric one encoded in the 3D Graph (i.e., proximity to windows, proximity to floors) to perform risk prediction assessing the formation of dew, which is often an early warning for more severe issues like indoor leaching or biological growth. More information about the 3DOnt Framework is available at: https://3dom.fbk.eu/projects/3DOnt.