3D reconstruction of semantic-rich digital twins for ACMV monitoring and anomaly detection via scan-to-BIM and time-series data integration

Abstract

Current research in air-conditioning and mechanical ventilation (ACMV) operation focuses on isolated sub-processes and analytical models. Digital twins, as digital replicas of assets, processes, or systems in the built environment, enable facilities manager (FM) to gain insights into the physical features of space, equipment performance, and energy efficiency. This study presents the 3D reconstruction of semantic-rich digital twins, which encompasses conditional and machine learning-enabled monitoring with 3D geometric models, for ACMV modeling and operation. The proposed framework involves a hybrid rule-based and data-driven approach to forecast the performance of indoor environment and identify potential anomalies throughout ACMV operation. Following this, a scan-to-BIM process is undertaken, with the aid of Simultaneous Localization and Mapping algorithms, to semi-automatically generate the as-built geometric models. Lastly, semantic enrichment of BIM is performed by incorporating time-series data from the rule-based and data-driven approach with 3D geometric models. The proposed approach supports the reconstruction of content-aware and semantic-rich digital twins, which utilize sensor-derived time-series data and 3D geometric models, to conduct advanced analysis for intelligent ACMV operation towards energy efficiency and occupant comfort.