Robotic part manipulation for 3D-print post-processing

• Main Author: Lim, Joyce Xin Yan
• Other Authors: Pham Quang Cuong
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Robotics; Additive manufacturing
• Online Access: https://hdl.handle.net/10356/174570
• Institution: Nanyang Technological University

The rise in Additive Manufacturing (AM) comes with unique opportunities and challenges. 3D-Print (3DP) allows rapid changes to part design, mass production, and massive part customization in manufacturing, which meets industrial manufacturing needs for customized parts such as dental moulds, shoe insoles, or engine vanes in turbo-machinery. However, a major drawback to 3DP that prevents its wide application stems from the bottleneck in post-production processes. Post-production tasks rely heavily on manual labor, which is tedious, repetitive and exposes the operators to hazardous substances. Some examples of post-processing are part cleaning, painting, sorting, and packing. Therefore, it is desirable to introduce robotics and automation in 3DP post-processing. However, current automated post-processing solutions are scarce and limited to specific materials or tasks. The opportunity for massive part customization also comes with unique challenges for the existing production paradigm of robotics applications. Two main challenges are the unique environment due to the presence of powder in AM, and the production of non-standard, geometrically complex parts due to customization. Hence, there is a need to develop generalized robotics solutions for implementation in end-to-end 3DP post-processing. As there are many AM technologies, this dissertation focuses only on powder-based AM processes. First, to demonstrate the feasibility of robotics in automated part cleaning, we developed a fully functional robotic prototype for the automated removal of residue powder on 3DP parts that mimics part cleaning by a human with a brush. Second, to support robot perception in 3DP post-processing, we proposed a fully automated vision pipeline for deep classification and localization of parts covered in powder, which is the first method that artificially simulates unique 3DP powder accumulation on the objects. Third, to support robot grasping and manipulation for batch-produced customized 3DP parts, we present an automated gripper customization method that designs versatile gripper fingers to grasp and manipulate a batch of objects resting at various positions with high precision. Fourth, to support part identification, grasping, and manipulation of unique parts with similar features, we introduce a method of pattern augmentation on 3DP parts, which has never been considered before, to perform grasping, part identification, and pose refinement in one-shot with a tactile gripper. With these contributions, we also sketched some possible directions for advancing the implementation of robotics and automation for 3DP post-processing.