Building Simulation 2027The BS2027 student modelling competition challenges participants to detect and diagnose operational faults in a mixed-use university building located in Graz, Austria. Entrants receive detailed building information, weather data, and two time-series datasets: one representing normal operation of the building for calibrating their simulation models, and another containing intentionally introduced faults. Using a validated building simulation tool, participants are expected to develop a simulation model of the building and use this model to identify fault occurrences temporally and spatially, diagnose their probable causes, and reconstruct a corrected “healthy” dataset describing normal building operation.
Machine-learning or AI-based approaches may be used as complementary methods to support the simulation-based analysis. Submissions will be evaluated based on the assumptions and methods used to develop the building simulation model, the precision of fault detection (temporal accuracy, spatial precision, and credibility of the source diagnosis), the quality of the reconstructed dataset, and the overall report. The use of machine-learning methods alongside physics-based models is highly encouraged and will be viewed positively in the evaluation process. Submissions can be individual or team-based, and all entrants must be enrolled as students (master or PhD) at the time of submission. Finalists will be invited to present their work at the Building Simulation 2027 conference in Vienna, including a poster and a short presentation.
