Koca, Ahmet Selim
(2025)
Detecting microscale surface impurities in powder bed fusion through light scattering and machine learning.
PhD thesis, University of Nottingham.
Abstract
Laser beam powder bed fusion (PBF-LB) is a leading metal additive manufacturing technique capable of producing complex components with high precision. However, the process is prone to forming microscale surface impurities such as balling, spatter, and surface pores, which can affect mechanical properties, surface finish, and overall part reliability. Detecting these impurities, particularly during fabrication, remains a significant challenge due to limitations in current inspection methods.
Although ex-situ techniques provide high-resolution analysis, they are often time-consuming, expensive, and/or unsuitable for real-time monitoring. In contrast, in-situ sensing approaches, while promising, frequently suffer from limited resolution, inconsistent accuracy across different machines, and a lack of reliability in detecting microscale impurities. Moreover, the real-time implementation of such methods is hindered by limitations in data acquisition, reconstruction, and analysis speed.
This thesis presents a novel method for detecting microscale surface impurities, which combines light scattering with machine learning (ML) to address these challenges. An optical setup was developed to capture light scattering patterns from PBF-LB surfaces. Experimental results demonstrated that surfaces with and without microscale impurities produce distinct scattering signatures. These signatures were subsequently used to classify the surfaces using ML algorithms.
To reduce the dependency on extensive experimental datasets, a simulation framework was established. Within this framework, light scattering patterns were generated using a model based on Beckmann-Kirchhoff (BK) theory. The surfaces input into this model were synthetically created using a generative adversarial network (GAN), enabling the generation of a large and diverse dataset. This approach considerably reduced the time and effort required to develop robust classification models.
The ML algorithms, once trained, achieved an accuracy of over 90% in detecting surface impurities. The system operates efficiently and possesses the potential for integration into a PBF-LB machine, enabling real-time, in-situ defect detection without interrupting the build process.
This study presents the first application of simulated light scattering patterns derived from synthetically generated surfaces detecting microscale impurities on PBF-LB surfaces. It demonstrates a rapid, cost-effective, and highly accurate method for assessing surface quality. Although currently implemented ex situ, the system architecture is well-suited for future integration as a real-time, in-process monitoring solution, thereby supporting more reliable and intelligent additive manufacturing workflows.
| Item Type: |
Thesis (University of Nottingham only)
(PhD)
|
| Supervisors: |
Leach, Richard
Hooshmand, Helia
Liu, Mingyu |
| Keywords: |
Laser beam powder bed fusion; Additive manufacturing; Microscale surface impurities; Light scattering signatures; Simulated light scattering patterns; Surface quality |
| Subjects: |
T Technology > TS Manufactures |
| Faculties/Schools: |
UK Campuses > Faculty of Engineering
UK Campuses > Faculty of Engineering > Department of Mechanical, Materials and Manufacturing Engineering |
| Item ID: |
82825 |
| Depositing User: |
KOCA, AHMET
|
| Date Deposited: |
09 Dec 2025 04:40 |
| Last Modified: |
09 Dec 2025 04:40 |
| URI: |
https://eprints.nottingham.ac.uk/id/eprint/82825 |
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
