Natural disasters strike anywhere, disrupting local communication and transportation infrastructure, making the process of assessing specific local damage difficult, dangerous, and slow. The goal of Building Damage Assessment (BDA) is to quickly and accurately estimate the location, cause, and severity of the damage to maximize the efficiency of rescuers and saved lives. In current machine learning BDA solutions, attention operators are the most recent innovations adopted by researchers to increase generalizability and overall performances of Convolutional Neural Networks for the BDA task. However, the latter, nowadays exploit attention operators tailored to the specific task and specific neural network architecture, leading them to be hard to apply to other scenarios. In our research, we want to contribute to the BDA literature while also addressing this limitation. We propose Global Voxel Transformer Operators (GVTOs): flexible attention-operators originally proposed for Augmented Microscopy that can replace up-sampling, down-sampling, and size-preserving convolutions within either a U-Net or a general CNN architecture without any limitation. Dissimilar to local operators, like convolutions, GVTOs can aggregate global information and have input-specific weights during inference time, improving generalizability performance, as already proved by recent literature. We applied GVTOs on a state-of-the-art BDA model and named it GVT-BDNet. We trained and evaluated our proposal neural network on the xBD dataset; the largest and most complete dataset for BDA. We compared GVT-BDNet performance with the baseline architecture (BDNet) and observed that the former improves damaged buildings segmentation by a factor of 0.11. Moreover, GVT-BDNet achieves state-of-the-art performance on a 10% split of the xBD training dataset and on the xBD test dataset with an overall F1- score of 0.80 and 0.79, respectively. To evaluate the architecture consistency, we have also tested BDNet’s and GVT-BDNet’s generalizability performance on another segmentation task: Tree & Shadow segmentation. Results showed that both models achieved overall good performances, scoring an F1-score of 0.79 and 0.785, respectively. / Naturkatastrofer sker överallt, stör lokal kommunikations- och transportinfrastruktur, vilket gör bedömningsprocessen av specifika lokala skador svår, farlig och långsam. Målet med Building Damage Assessment (BDA) är att snabbt och precist uppskatta platsen, orsaken och allvarligheten av skadorna för att maximera effektiviteten av räddare och räddade liv. Nuvarande BDA-lösningar använder Convolutional Neural Network (CNN) och ad-hoc Attention Operators för att förbättra generaliseringsprestanda. Nyligen föreslagna attention operators är dock specifikt skräddarsydda för uppgiften och kan sakna flexibilitet för andra scenarier eller neural nätverksarkitektur. I vår forskning bidrar vi till BDA -litteraturen genom att föreslå Global Voxel Transformer Operators (GVTO): flexibla attention operators som kan appliceras på en CNN -arkitektur utan att vara bundna till en viss uppgift. Nyare litteratur visar dessutom att de kan öka utvinningen av global information och därmed generaliseringsprestanda. Vi tillämpade GVTO på en toppmodern CNN-modell för BDA. GVTO: er förbättrade skadessegmenteringsprestandan med en faktor av 0,11. Dessutom förbättrade de den senaste tekniken för xBD-testdatauppsättningen och nådde toppmodern prestanda på en 10% delning av xBD-träningsdatauppsättningen. Vi har också utvärderat generaliserbarheten av det föreslagna neurala nätverket på en annan segmenteringsuppgift (Tree Shadow segmentering), vilket uppnådde över lag bra prestationer.

Tunnelling in urban areas causes short-term and long-term ground movements under existing buildings. Finite element analysis provides a useful option for assessing the likely extent of damage induced in these buildings. Although finite element analysis is suggested to be used in the later phases of the building damage assessment procedures employed in practice, only the effect of short-term ground movements is typically considered and there are no detailed guidelines on the specification and complexity of the modelling. This thesis addresses the tunnel-soil-building interaction problem and the effect of long-term consolidation, as well as demonstrating the application of 3D finite element analysis with appropriate simplifications for practical assessment purposes. Finite element models are developed to quantify the effect of shallow tunnelling on an example masonry building founded on strip footings, considering both single- and twin-tunnel scenarios in a typical London soil profile. Total stress and effective stress analyses are adopted with specific modelling procedures to focus on the short-term and long-term response respectively. The analyses use a non-linear model for the masonry, and allow slippage and gapping at the soil-footing interface. Two advanced constitutive models for the soil (the extended Mohr-Coulomb model and the modified two-surface kinematic hardening model) are implemented with customized stress update schemes. The finite element results present the interaction between the soil and the building by comparing with the greenfield ground response. The horizontal coupling between the foundation and the ground is shown to be relatively weak. The dominant deformation mode of the building varies with the tunnel configuration (i.e. single or twin tunnels) and the tunnel eccentricity. Strain localization is found around the explicitly modelled window and door openings. The long-term consolidation is sensitive to the permeability of the tunnel lining. The building response to long-term ground movements is further affected by the tunnel-tunnel interaction in the case of twin-tunnel configuration. Performing 3D analysis of a single facade and foundation provides useful damage predictions, without the need to model a complete building. The proposed result processing methods such as characteristic strain and damage bar chart are practical tools for assessment. The study highlights some limitations of the elastic beam assessment method, which is often adopted in the early phase of the damage assessment process.

L’objectif de cette thèse est, d’une part, de proposer une amélioration des méthodes d’estimation des cuvettes d’affaissement et des méthodes d’évaluation des dommages susceptibles de se produire sous leurs effets et de l’autre, de développer des outils basés sur ces méthodes pour étudier les affaissements et les dommages sur des cas pratiques. L’étude de l'influence de la topographie sur les cuvettes d'affaissement dans des conditions d’exploitation simplifiées grâce à des modèles numériques avec des profondeurs d'exploitation et des pentes du sol variables a permis de proposer une nouvelle fonction d’influence basée sur une densité de probabilité normale asymétrique lorsque la surface du sol est non-plane. Une modélisation simplifiée des habitations en maçonnerie sous la forme de deux modèles de structures bidimensionnels croisés, alignés avec les axes d’inertie de la structure étudiée et dans lesquels la méthode des déplacements est mise en œuvre pour calculer les efforts internes et les déformations sous l’effet de déplacements imposées des fondations. Ces modèles simplifiés dont les caractéristiques géométriques et mécaniques sont définis pour chaque type de bâtiment étudié permettent d’estimer les efforts appliqués à chaque bâtiment d’une ville exposée à un affaissement de terrain et de fournir de nouveaux critères d’évaluation des dommages prenant en compte davantage d’informations que les méthodes habituelles. Une estimation des dommages dans la ville de Joeuf sur la base des nouvelles méthodes proposées, tant pour le calcul de l’affaissement que pour l’estimation des dommages, a été réalisée / The objective of this thesis is to improve the methods of subsidence computation and building damage evaluation, and to develop some tools based on these methods to study the mining subsidence and building damage cases in Lorraine. By investigating the topography influence on subsidence under simplified mining conditions, and using numerical models with varying mining depths and ground surface angles, a new influence function method, which is based on a probability density function of a skew normal distribution, to simulate the element subsidence, was firstly developed and can be used to compute the mining subsidence caused by the excavation under non-flat surface. Then, plane framed structural models were chosen to study the mechanical behavior of 3D buildings. For each building, two plane models located in the vertical sections passing through the principle inertia axes of the building’s projective polygon were considered. Their geometry and mechanical characteristics were chosen according to the construction type and used materials of the building under consideration. Then, by using the matrix displacement method with some modifications, the internal forces and displacements for the entire structure could be computed. The achieved internal forces could then be compared to damage grade criteria to determine the extent of building damage.Finally, by using the improved methods of subsidence computation and building damage evaluation, a real case application was performed in Joeuf city (France). The subsidence was computed and applied to the defined structural models as support displacements, and then the damage extents of the buildings in Joeuf were predicted