Research

Research Focus

My PhD research focuses on the development of the Digital Twin (DT) concept, which integrates both mathematical models and data-driven approaches for monitoring and managing the health of civil engineering structures. By combining the Reduced Basis (RB) method and Deep Learning (DL) technology, my research aims to detect abnormal changes in structural behaviour in real-time, allowing for immediate interventions to prevent potential failures. This innovative approach addresses uncertainties in physical models and provides a synchronised interaction between physical and virtual models using IoT technology.

The primary aim of this research is to develop a real-time damage detection system that bridges the gap between theoretical models and real-world structural conditions. By leveraging the DT framework, my work has demonstrated significant improvements in the accuracy and efficiency of structural health monitoring (SHM). The findings include:

• Development of real-time structural anomaly detection using deep learning algorithms.
• Introduction of RB methods to reduce the computational complexity of large-scale models, ensuring faster response times.
• Successful integration of IoT systems for continuous, real-time monitoring of physical assets.
• A solution for managing model uncertainties, improving the reliability of predictive maintenance strategies in civil infrastructure.

Research Interests

My current research interests aim to expand my knowledge in various areas related to SHM, DTs, and advanced machine learning applications.

• Generative AI and Physics-Informed Models: I am researching the application of generative AI in creating more accurate and reliable DTs. By integrating physics-informed models, my work aims to bridge the gap between theoretical predictions and real-world behaviour in SHM. Additionally, my research explores how generative AI can be used to visually represent structural damage in real-time within the DT environment.
• Internet of Everything (IoE) and SHM Integration: Expanding upon traditional IoT concepts, I explore the integration of the IoE with SHM. This research focuses on incorporating human insights alongside data from machines and systems to create a more comprehensive approach to monitoring, data collection, and real-time analysis.
• Self-Generating Digital Twins: My passion lies in the development of self-generating digital twins for SHM. These innovative DTs can autonomously update and improve based on real-time data, allowing for continuous adaptation to the changing conditions of infrastructure, reducing the need for manual updates.
• Self-Healing Materials: I am investigating the potential of self-healing materials in SHM systems, particularly bacterial concrete for concrete structures and shape memory alloys for steel structures. These materials, capable of autonomously repairing damage, contribute to the autonomous lifecycle of infrastructure, complementing the self-generating digital twins by extending the longevity and resilience of structures without human intervention.
• Technological Focus: My technology interests align closely with my research in SHM. I am particularly focused on graph diffusion models, dynamic diffusion models, and improving the efficiency of machine learning algorithms used in SHM and digital twin development. I am also exploring Large Language Models (LLMs), knowledge graphs, and LLMs enhancement through Retrieval-Augmented Generation (RAG) and Masked Augmented Generation (MAG) models.

Ongoing Learning Initiatives

Revisiting Civil Engineering Curriculum for Machine Learning Insights: I am actively revisiting all of the courses I studied during my Bachelor's degree in Civil Engineering. This personal exercise involves a comprehensive comparison of the notes I originally took during my studies with the new notes I am currently writing. The goal is to reflect on how my understanding of civil engineering concepts has evolved over time and how these concepts can now be applied to machine learning and modern technological advancements.

I am revisiting core subjects such as:

• Structural Mechanics
• Fluid Dynamics
• Geotechnical Engineering
• Materials Science
• Environmental Engineering
• Transportation Engineering

Through this ongoing project, I am transforming my original undergraduate notes into a living knowledge resource that bridges classical engineering principles with the capabilities of AI, machine learning, and Digital Twins. The following aspects define this process:

• Comparative Knowledge Maturity: I systematically compare my past academic notes with my current research-informed understanding, reflecting on how traditional engineering methods can evolve with advancements such as physics-informed machine learning and real-time data-driven modelling.
• Integration of Machine Learning and Physics-Informed Techniques: I reframe key engineering concepts by embedding modern computational perspectives. For example, classical structural analysis methods are augmented by exploring how Physics-Informed Neural Networks (PINNs) and Dynamic Diffusion Models can enhance real-time structural health monitoring within Digital Twins.
• Application Mapping to Self-Generating Digital Twins: I actively explore how each engineering discipline contributes to the development of autonomous, self-updating Digital Twins. This includes identifying how machine learning algorithms, generative AI, and IoT data streams can complement traditional engineering models to create predictive, resilient, and self-generating infrastructure systems.

• Comparing Past and Present Understanding: I compare the notes I took during my initial studies with the new notes I am currently compiling. This process enables me to assess how my understanding of complex engineering principles has matured. My focus is on how traditional methods can be re-evaluated in light of modern advancements in computational techniques, particularly machine learning.
• Rewriting Notes with Machine Learning Perspectives: As I write my notes, I incorporate new insights gained from my research in machine learning and DTs. This process highlights the potential of machine learning in improving the accuracy and efficiency of engineering models. For example, traditional methods of structural analysis can now be enhanced using physics-informed neural networks (PINNs).
• Linking Engineering Concepts to Machine Learning Applications: I identify parallels between traditional engineering models and machine learning algorithms. For instance, fluid dynamics, which traditionally relies on physical simulations, can be compared to machine learning approaches that simulate and predict flow behaviour in real-time. My revisited notes now reflect these connections, enhancing my understanding of how machine learning can be applied to physical systems.
• Continuous Learning and Reflective Practice: The ongoing nature of this project reflects my commitment to continuous learning. Revisiting my previous courses allows me to update my knowledge and reflect on how new technologies and machine learning can offer solutions to longstanding civil engineering challenges. This exercise also serves to sharpen my skills in translating domain-specific knowledge into data-driven frameworks.

I consider this project an essential part of my ongoing professional development. Combining the original notes with new insights from my work on DTs and machine learning has allowed me to develop a deeper understanding of both civil engineering and the computational techniques that are revolutionising the field.