The goal of the META-NOVIB project is to develop a comprehensive framework to effectively predict and control the vibration and noise induced by underground railway tunnels using digital twin technology supported by machine learning tools. This system provides valuable insights for engineering decisions throughout the operation and maintenance of these tunnels. Additionally, it evaluates the performance of seismic metamaterials (SMM) in attenuating the level of noise and vibration to meet the allowable limit. META-NOVIB will provide an integrated platform for visualisation and real-time prediction and virtual control of the railway-induced noise and vibration during the operation and the maintenance phase. Thus, the output will have wide implications on the health of nearby residents due to vibrations and prevent any structural damage to historical buildings or structures, with high academic and industrial impact.

Leveraging the power of a computational framework based on CutFEM combined with a BIM platform incorporating CAD-based data, the TwinSSI project will develop a comprehensive DT for underground design and construction. To validate the computational framework, real-scale experiments of tunnel-soil-structure interaction will be performed at the National Buried Infrastructure Facility (NBIF) at UoB. Moreover, the developed DTs will be applied to real case studies co-created with the industrial partners Network Rail and Maidl Tunnelconsultants. The TwinSSI project will thus, for the first time, create and validate detailed DTs in the domain of soil-structure interaction modelling. The project outcomes will lead to a new paradigm for project planning and monitoring by

geotechnical engineers.

Project is grounded in the realm of underground spaces and aims to delve deeply into the concept of highly automated digital twinning and its seamless integration with comprehensive simulation. It highlights the innovative application of digital twin technology in the context of underground spaces, encompassing highly synchronized modelling of underground structures and their built environments. This includes the acquisition of structural dimensions and semantics, real-time defect monitoring and projection, and precise description and simulation of geological surroundings. A visual interactive platform that allows users to query and interact with these digital twins will be developed.

This research aims to revolutionize structural defect management in railway tunnels by leveraging advanced automation and machine learning. This project aims to develop an automated, prompt-enabled system for recognizing and categorizing structural surface defects from ATE/DIFCAM data using SAM-based instance segmentation models. By automating defect detection, the proposed approach enhances the accuracy and consistency of defect identification while significantly reducing the time and effort required for manual interpretation and classification processes. Additionally, the system ensures compliance with established standards, including Network Rail's Standard 006 and GIS-based condition monitoring systems such as WebTCMI, fostering seamless integration into existing workflows.

The M-Twin4US project aims to develop an innovative Digital Twin (DT) platform for the maintenance of buried pipelines, with potential for broader underground infrastructure applications. Combining advanced sensing technologies such as Ground-Penetrating Radar (GPR) and closed-circuit television (CCTV) with state-of-the-art machine learning methods like Multi-task Transformers and Segment Anything Models (SAM), the project addresses critical challenges in soil-pipe interaction, GPR response prediction, and condition assessment. By integrating digital modeling, machine learning, numerical simulations, and real-scale testing, M-Twin4US will create a customizable surrogate model for predicting pipeline performance and offer an AI-powered toolkit to help stakeholders make proactive maintenance decisions.

Project is dedicated to constructing and improving the workflow of Scan-to-BIM-to-FEM. Our aim is to find an effective and feasible method to automatically reconstruct scanned structures (like bridge) information (images or point cloud data) into semantically enriched 3D parametric and updatable models that support functionalities such as defect detection. Through further extraction and processing of key information from the models, we import them into finite element analysis software (such as ABAQUS) for numerical simulation, enabling tasks such as safety assessments, deformation detection, and structural optimization. Moreover, the results of the simulation can be effectively interacted and visualised with the BIM software. Ultimately, we will achieve an integrated workflow for the detection, evaluation, and optimization of structures. 

The project focuses on the development of a scalable framework for 3D reconstruction of large outdoor environments using Neural Radiance Fields (NeRF).  This research aims to address key challenges such as handling massive datasets, environmental variations (e.g., lighting and weather), and the integration of deep learning for improved scene understanding. By employing distributed computing, appearance encoding for lighting consistency, and incremental learning, the project seeks to improve modeling accuracy and efficiency. Furthermore, the integration of NeRF with BIM, and VR technologies opens possibilities for innovative applications in construction management, semantic segmentation, and real-time interaction with virtual models, significantly enhancing automation in intelligent construction processes.

The project focuses on developing AI-supported structural health monitoring (SHM) systems to enhance safety and maintenance efficiency in buildings and infrastructures. The goal is to create advanced damage detection methods and data augmentation techniques, allowing efficient use of existing datasets through data recycling. A key objective is to design a model capable of identifying a variety of structural issues, such as cracks, deformations, and weaknesses, while integrating additional data sources like sensor data (e.g., temperature, humidity, soil conditions) and historical maintenance records using federated learning. The model will also predict potential outcomes of detected damage, including structural failure risks and timeframes, while providing real-time monitoring and actionable insights through an intuitive user interface, aiding building owners and maintenance teams in decision-making.

This project aims to develop novel method for creating Digital Twins of existing estates through an ML-enabled automatic modelling process using point cloud data. By using point cloud data, this project seeks to automate the model creation process, ensuring greater accuracy and semantically enriching the model  while reducing the model reconstruction time. To address the above problem, ML algorithms will be designed  to analyse and learn from architectural design patterns across various periods. This ML-driven approach will allow the system to infer missing MEP details and intelligently link them based on partial information. The program will be capable of making reasonable assumptions to fill in these gaps, thereby producing a comprehensive digital representation of the estate. By automating this process, this project aims to create a reliable, AI-powered platform that makes digital estate modelling more accessible for estate management.

The main goal of Ali Gamra's research is to predict building damage due to tunnelling activities using FEM-trained, ML-based models integrated into a BIM framework. Tunnelling induces highly nonlinear behaviour, and predicting building damage is a challenge, especially when multiple scenarios with varying input parameters are considered. He employs the Volume Loss (VL) method to estimate settlements, uses the Euler Bernoulli Beam of Two Parameter Elastic Foundation (EBBEF2p) for SSI, and merges all components in FEM, including the use of the Concrete Damaged Plasticity (CDP) Model for the non-linear assessment of buildings,  and conducting hundreds of simulations. The resulting dataset of inputs and outputs is then used to train ML models. By integrating these models into BIM via a Python-based plugin, near real-time predictions of building damage are enabled. This solution eliminates the need for prior expertise in FEM, ML, or SSI, making it highly accessible. The impact is significant in scenarios requiring multiple tunnel design alternatives, considering various parameters and building configurations. The highly accurate ML models, trained on diverse datasets, provide a flexible and efficient tool for safe and optimised tunnel design.

This project aims to develop a cost-effective method for creating Digital Twins of existing estates through an automatic modelling process using point cloud data. The core focus of the project is to reduce the high cost typically associated with generating digital models of existing MEP systems. By using point clouds, this project seeks to automate the model creation process, ensuring greater accuracy while reducing the time and financial resources needed. To address the above problem, AI will analyse and learn from architectural design patterns across various periods. This AI-driven approach will allow the system to infer missing MEP details and intelligently link them based on partial information. The program will be capable of making reasonable assumptions to fill in these gaps, thereby producing a comprehensive digital representation of the estate. By automating this process, this project aims to create a reliable, AI-powered platform that makes digital estate modelling more accessible for estate management.

The project is targeted at developing a highly automated and comprehensive framework for GIM-to-FEM undeground modelling, by seamlessly integrating digital ground information models into the FEM numerical simulations. To account for soil spatial variability and multi-level uncertainty, as well as reflect up-to-date in situ state , GIMs are intented to feature real heterogeneous geological conditions and conditionally distributed geotechnical parameters, updated with new field investigation data. Strenghtening the probabilistic foundations and streamlining the data exchange interoperability, GIM-to-FEM will offer an advanced and efficient approach to fusing digital and numerical models for enhanced precision in undeground costruction and design problem solving.

The OptiMACS project aims to optimize the design of modern aeronautical composite structures by developing and implementing advanced Multidisciplinary Design Optimization (MDO) tools for the European aerospace industry. Addressing various operational constraints like reliability, strength, and manufacturability, the project will enhance the accuracy and efficiency of AIRBUS's existing MDO platform. OptiMACS will also train the next generation of MDO professionals, providing a supportive environment for five Early Stage Researchers (ESRs) to develop their research and transferable skills. The initiative combines expertise from mechanical, aerospace, manufacturing, software engineering, and applied mathematics, fostering innovation in both academia and industry. 

The No2Noise Training Network addresses the urgent need for lightweight and multifunctional composite structures in modern transport applications, which often suffer from poor dynamic and acoustic isolation. To overcome these challenges, the project focuses on developing a high-fidelity Multidisciplinary Design Optimization (MDO) scheme for composites with poroelastic inclusions, aiming to balance minimal mass with maximum damping and comfort. Research will involve multiscale modeling to understand acoustic wave interactions with complex materials. Additionally, N2N offers a supportive environment for three Early Stage Researchers (ESRs), fostering both research and transferable skills in multidisciplinary settings. 

This project developed a novel Machine Learning (ML) based approach for fusion of numerical and experimental data for the prediction of the extended hollo-bolt connection behavior in terms of strength, stiffness, and column face displacement, as well as characterisation of the failure mode. We propose two strategies for data and model fusion to leverage the predictive capabilities and scalability of both experimental and numerical data sets. To promote wide use of the proposed meta model, we developed a graphical user interface as a standalone program.  The developed program for characterisation of EHB connection behaviour in tension is openly available for wide use by the civil engineering community. 

Multidisciplinary design optimisation (MDO) is a tool commonly used in aircraft design. The procedure relies on a global FE-model (GFEM), which does not contain a detailed representation of the structure geometry.

At the same time, aircraft structures exhibit at multiple locations features with a non-regular geometry, which require detailed FE-modelling (DFEM). Including them in the multidisciplinary optimisation would result in a prohibitive computational cost, therefore they are ignored.

In order to account for the influence of non-regular parts on the optimisation, at an acceptable computational cost, this research project develops a global-local methodology to extend an MDO procedure for aircraft composite structures.

The SATBIM project has developed a multi-level simulation model for tunnel-structure interaction integrated in the framework of Building Information Modelling to support engineering decisions during the project life cycle and to allow for the evaluation and minimization of risks on the existing infrastructure. SATBIM is an integrated platform for structural analysis, visualisation and optimization of the mechanized tunnelling process from early stages of the design over to the construction and the operation phase. The complete concept will be validated using industrial data with reference to the Rastatter tunnel project in Germany. The output will have wide implication on technology with expected high academic and industrial impact.

Dr Ninic's thesis presents a comprehensive approach to simulating the shield tunneling process using a process-oriented Finite Element (FE) model. The first part describes the model's components, focusing on time-dependent effects such as consolidation and grouting material properties, along with the analysis of loading on the tunnel lining and soil-structure interactions. The second part introduces an embedded beam formulation for analyzing interactions between pile foundations and surrounding soil, applicable to shield-driven tunnel construction near existing buildings. Finally, a real-time steering method for mechanized tunneling is proposed, utilizing meta models, Artificial Neural Networks (ANNs), and Particle Swarm Optimization (PSO) to optimize steering parameters based on monitoring data from the Wehrhahn metro line project in Düsseldorf, Germany