AI Enabled 3D Optical Sensing System for in-process High Dynamic Additively Manufactured Surface Inspection

The proposed project aims to address key problems related to a previously developed in-process surface measurement deployed in an EBeam Metal Additive Manufacture (AM) machine. AM is a developing process in industry that produces complex geometry parts by a layer by layer melting and solidification process of metal powder. Specifically, the project aims to develop an intelligent optical fringe projection system that can dynamically adjust optical measurement parameters based on the dynamic processing of the measurement surface. The developed technical method will be applied to a commercially AM machine currently marketed by the project partner. In the process of EBeam AM, metal powder is melted at high temperature, fused, and then solidified to become a shiny metal surface. Consequently, the texture, shape, reflectivity, curvature, and other characteristics of the printed surface have changed compared with metal powder. The fringe projection, surface layer, measurement system needs to address the dynamic range surfaces measurement issues.
Therefore, the following Work Packages will be implemented,

WP1) Dataset acquisition: A large training dataset of surface images with varying characteristics will be acquired using the fringe projection system. The samples will be manufactured by the AM machine at project partner Wayland Additive.
WP2) Pre-processing: The acquired images will be pre-processed to remove noise and enhance the image quality. The pre-processing techniques will include denoising, filtering, and lens distortion corrections.
WP3) Feature extraction: Relevant layer geometry features will be extracted from the pre-processed images to train machine learning models. The features to be extracted include surface texture, reflectivity, projector intensity, exposure time all of which reflect the manufacturing features such as melting speed and melting power.
WP4) Investigate machine learning algorithm models: Various machine learning algorithms, such as decision trees, support vector machines, and deep learning, will be trained using the extracted geometry features from the acquired images. The algorithms will be evaluated based on their performance in predicting the surface characteristics.
WP5) Adaptive measurement system: The trained algorithms will be used to predict the optimum measurement parameters based on the surface characteristics. The adaptive system will be updated in real-time to ensure accurate and robust 3D surface measurement.

This project has the potential to be an exciting and impactful area of research. Developing more advanced detection systems for additive manufacturing will help improve the accuracy and efficiency of the manufacturing process and reduce waste or errors. Additionally, it could enable researchers and manufacturers to more quickly develop and test new materials or printing techniques. The proposed research project will be conducted across both the company(Wayland Additive) and the university (UoH). Initially, prototype development and algorithm research will be carried out at UoH, with samples processed by the AM machine (Wayland) used for testing. Following this, the algorithms will be debugged using the commercial machine. Once they have passed testing, the algorithms will be finally applied in the commercial AM machine. The proposed method has the potential to solve a crucial inspection problem in additive manufacturing processes, enhance manufacturing efficiency, overcome limitations in threedimensional measurement technology, deepen research, and increase the practicality of measurement technology. This method can facilitate the transformation of academic research into industrial applications. The funding requested covers seconded Researcher salary costs of the PI and a range of optics consumables.