Understanding conservation condition and rehabilitation needs at district scale in time of energy saving and climate change

NTNU Norges Teknisk-Naturvitenskapelige Universitet (NTNU) Norway

PhD enrolment : NTNU

Supervisor: Chiara Bertolin

Co-supervisor: Chao Gao

Start date September 2026

Duration (in months) 36

Objectives: This project aims to develop deep learning neural networks to investigate data in available database (e.g., EU building stock observatory and other similar database) to (i) classify building stock at district scale more prone to decay based on buildings construction year, location, main constitutive material, exposure, use and to (ii) predict potential maintenance and restoration priority based on (i) and existing energy retrofit directive and/or risk maps.

Expected Results: (i) to build high-quality dataset highlighting the façade that is more prone to decay at district/city level from existing raw database; (ii) to apply deep learning neural networks (e.g., convolutional neural network) to perform semi-automatic decay pattern segmentation and associated mapping; (iii) to correlate the maintenance of buildings at district/city level with decay patterns/features at district scale to prioritize large scale intervention; and (iv) to optimize the maintenance activities by leveraging these correlations learned by machine-learning based approaches.

Planned secondment(s): FORTH , P. Fafalios, M12, 4 weeks, AI and ML algorithms, CY (Pau) K. El Ganaoui M24, 4 weeks, AI materials