Metrology, the scientific study of measurement, is devoted to enhancing manufacturing performance and controlling various functions on engineering components by boosting speed, accuracy, as well as lowering the cost of measurement. Surface metrology is a major branch aiming at measuring and characterising surface topography and features. Therefore, adequate surface characterisation is crucial for investigating and understanding surface properties for optimising manufacturing processes and improving functionality control. It is defined as extracting and quantifying surface topography through the examination of specific features and patterns using a set of parameters that reflect the quality of the surface. These parameters are believed to be able to interpret functional attributes such as the optical quality, durability, and reliability of the surface. In the context of contemporary manufacturing technologies such as high-precision technology, additive manufacturing, and computer technology, engineering surfaces are growing more intricate with many deterministic patterns. Consequently, the ability to adequately assess these complex freeform structured surface geometries can help optimise the performance and attain meticulous control of such specialised functional components. Modern metrology is transitioning towards digitalised and intelligence-enabled systems. One primary challenge in future development is to create a sophisticated data analytics system that can be incorporated into the manufacturing process. In particular, surface metrology will enhance our comprehension of complex systems through digital surface texture analysis technology and serve to assist an intelligent decision-making framework for smart manufacturing over the next few decades. Therefore, this Ph.D. project aims to develop an intelligent, AI-assisted framework for the characterisation of complex freeform structured surfaces. The framework is grounded in standard surface characterisation processes—form removal, denoising, and segmentation with parameterisation—and leverages a range of advanced machine learning techniques in conjunction with computer vision-based image processing methods. These include supervised learning, unsupervised learning, deep neural networks, and physics-informed machine learning. These approaches are explored and applied to enable automated, accurate, and intelligent surface analysis. The thesis is organised into six chapters: Introduction, Literature Review, Advanced Form Removal, Advanced Denoising, Advanced Segmentation}, and Conclusion and Future Work. The introduction outlines the background, motivation, research gaps, and objectives of the study. The literature review provides an overview of conventional methods in surface characterisation, tracing their evolution and identifying opportunities for advancement through machine learning. The three central chapters present the core research contributions, each corresponding to a key step in the surface characterisation framework. They include comparative studies of traditional techniques and propose novel learning-based methods supported by theoretical analysis and experimental validation. The final chapter synthesises the research findings and outlines future directions.
| Date of Award | 19 Jan 2026 |
|---|
| Original language | English |
|---|
| Sponsors | Engineering and Physical Sciences Research Council |
|---|
| Supervisor | Jane Jiang (Main Supervisor) & Wenhan Zeng (Co-Supervisor) |
|---|