In the heart of Milan, Italy, researchers at the Politecnico di Milano are revolutionizing the way we monitor and predict the dynamic behavior of heritage masonry buildings. Led by Alireza Entezami from the Department of Civil and Environmental Engineering, a novel approach to dynamic response prediction is being pioneered, with significant implications for the energy sector and beyond.
The challenge at hand is clear: effective dynamic monitoring of heritage masonry buildings relies on comprehensive data from multi-sensor systems. However, practical limitations often lead to incomplete datasets, posing a significant hurdle for accurate predictions. Entezami and his team have tackled this issue head-on, developing a dual-phase residual-augmented regression (DPRAR) method that integrates random forests (RF) with a deep regression-based neural network (DRNN).
The process is ingenious in its simplicity. Initially, the RF algorithm predicts responses from the available measured data, extracting residuals between observed and predicted values. These residuals, serving as latent information, are then combined with the measured data to form an enhanced dataset. This enriched data set is used to train the DRNN, which ultimately makes the final predictions.
“The beauty of this method lies in its ability to innovatively use residuals to address missing unmeasured factors,” Entezami explains. “By integrating statistical and deep learning regressors, we’ve created a robust tool that significantly improves dynamic behavior prediction, even with limited environmental measurements.”
The implications of this research are far-reaching, particularly for the energy sector. Heritage masonry buildings, often characterized by their unique architectural features and historical significance, require careful monitoring to ensure their preservation. Accurate dynamic response prediction can aid in optimizing energy efficiency, reducing maintenance costs, and preventing potential structural failures.
Moreover, the DPRAR method’s ability to work with incomplete data makes it a practical solution for real-world applications. As Entezami notes, “Our approach offers an intelligent solution to the practical limitations of in-situ measurements, making it a valuable tool for engineers and conservators alike.”
Published in the journal ‘Developments in the Built Environment’ (translated to English as ‘Developments in the Built Environment’), this research marks a significant step forward in the field of structural health monitoring. By addressing the challenges of data scarcity and integrating advanced machine learning techniques, Entezami and his team are paving the way for more accurate, efficient, and cost-effective monitoring of heritage buildings.
As the energy sector continues to evolve, the need for innovative solutions to monitor and maintain our built heritage becomes increasingly apparent. The DPRAR method developed by Entezami and his colleagues offers a promising avenue for achieving these goals, shaping the future of dynamic response prediction and structural health monitoring.