Time-Series Forecasting of Structural Temperature in Heritage Buildings Using Regression and Deep Learning Approaches
Waqas Qayyum, N. Cavalagli, E. García-Macías, F. Ubertini
Accurate prediction of the structural temperature field is crucial for the static and dynamic monitoring of engineering structures, with particular significance for heritage buildings where material preservation is paramount. The complex, time-lagged, and non-linear relationship between external air temperature and structural thermal response poses a significant challenge for traditional empirical or statistical methods. This study proposes a framework utilizing statistical regression and advanced recurrent deep learning models to accurately predict structural temperature based on external data. A key focus is the efficient generation of input features to capture delayed and cumulative thermal effects. The methodology is applied in a case study to predict temperatures at various points within a historic basilica. The primary objectives are to remove temperature-induced effects from structural monitoring data for a more accurate assessment of the building's performance and to provide a robust method for imputing missing data. A comparative analysis demonstrates the performance of the models, evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). The results offer a reliable and precise methodology for selecting the optimal approach for structural temperature field estimation and data correction in the assessment of heritage structures.