From Sensor Time Series to Building Performance Indicators: A BIM-Informed Framework for Contextual Data Integration

Abstract

The integration of Building Information Modeling (BIM), sensor data, and data-driven analysis can support a more contextual interpretation of building performance in existing buildings. This paper presents a pilot study on sensor-driven, BIM-informed building performance analysis, using the sensor network installed in the "F" building of the Budapest University of Technology and Economics as a case study. The aim is not to develop a complete digital twin platform, but to demonstrate how indoor environmental sensor measurements can be transformed from isolated time series into contextualized building performance information. The proposed approach combines sensor data preparation, quality assessment, sensor-to-BIM mapping, and exploratory analysis of indoor environmental patterns. Temperature and relative humidity measurements were processed to identify data quality issues, invalid measurements, outliers, temporal patterns, correlations, and clustering-based response groups. The sensor-to-BIM mapping linked measurement points to rooms, floors, functions, zones, spatial contexts, and BIM identifiers, enabling interpretation within spatial and semantic building context. The pilot results show that data quality has a direct influence on analytical outcomes, including clustering results. Contextualized sensor analysis supported the identification of stable rooms, strongly fluctuating spaces, and rooms with unusual environmental responses. A preliminary temperature prediction using a Long Short-Term Memory neural network for the F29 lecture hall also demonstrated the potential of short-term forecasting as a future extension. The study highlights that sensor-informed BIM can provide an intermediate layer between raw monitoring data and future digital twin-based operation.