Abstract
With the rapid growth of the manufacturing industry, laser-based metal additive manufacturing, such as laser powder bed fusion, has the potential to usher in a revolution. However, its widespread adoption is contingent on the resolution of several challenges. A significant challenge is the uncertainty associated with part consistency when standardized materials are used in additive manufacturing processes. To ensure the quality and reproducibility of AM parts, it is essential to ensure that consistency is maintained. This study delves into an assessment of part-to-part consistency, leveraging a Pyramid learning-based technique that utilizes X-ray computed tomography (XCT) images for four nominally identical parts. Employing machine learning, this approach adopts a hierarchical feature system to enhance model performance. Pyramid Learning not only improves the accuracy of part-to-part consistency scanning but also reduces noise, bolstering overall robustness. The findings showcased the efficacy of pyramid learning in enhancing performance metrics when sufficient detail is present. It also provides guidance on locating defects and deformations for the AM parts.
