Machine Learning Techniques for Assessment of Structural Damage in Resource-Constrained Environments (VC24091)

The University of the West of Scotland (UWS) is seeking to attract a PhD candidate of outstanding ability and commitment to join its vibrant and growing programme of internationally excellent research.

The use of computer vision-based AI models enables the efficient detection of structural faults in public infrastructure, aiding inspectors in performing structural health monitoring. This technology allows for timely interventions and predictive maintenance, enhancing the speed and accuracy of inspections.

This studentship focuses on the research, development, and practical application of machine learning methods to detect and quantify structural damage in surfaces such as concrete or asphalt using image or LiDAR data. The primary goal is to develop solutions suitable for resource-constrained environments, characterised by limited data or computing resources, often necessitating edge computing.

Addressing these challenges is crucial, as larger models and datasets typically demand substantial computing resources, leading to increased energy consumption, carbon emissions, and higher costs. The objectives include creating efficient new algorithms for real-time structural damage detection and quantification (e.g., cracks, spalling, corrosion, and exposed rebar). Additionally, there will be a focus on designing new semi-supervised and/or unsupervised techniques and addressing domain shift issues to counteract the necessity of large, annotated datasets. Ultimately, this would culminate in integrating these methods into standalone hardware systems.

Specifically, the student will design and use neural networks to process data, aiming to enhance the detection and quantification of structural damages. Systematic experiments will compare the proposed methods against state-of-the-art techniques using datasets from various domains (e.g., roads, bridges, tunnels, and buildings) while evaluating performance indicators such as speed and detection capability.

The outcomes of this project enable a more sustainable and cost-effective approach to visual infrastructure maintenance, ultimately reducing examiners’ exposure to potentially hazardous environments, and enhancing their safety. In turn, this will lead to increased safety and longevity of urban infrastructure, resulting in a positive social impact.

Candidates should have:

 –         A Master’s degree with a classification of Merit or higher, and/or an undergraduate degree with at least a UK 2:1. A Distinction or 1st class degree is highly desirable. Relevant fields include computer science, machine learning, data science, engineering, or any field demonstrating a strong computing and mathematical background.

–         Proficiency in Python and/or C/C++ is essential. Experience implementing and using deep learning or computer vision methods is desirable.

–         Strong mathematical skills and the ability to write clearly and effectively are required.

–         The ideal candidate should be self-motivated, eager to learn, and passionate about using data science and computer vision applied to real-world problems.

For more information about the PhD Studentship or any enquiries about the project, please contact Dr Jacob Koenig at jacob.koenig@uws.ac.uk

 Applications must be made via the UWS Online Application Link: www.uws.ac.uk/study/research-degrees/admissions-application/postgraduate-research-application-guide

Application Deadline: 09/11/2024

Start Date: 01/02/2025