Abstract
Depowdering is a vital post-processing step in Powder Bed Fusion Additive Manufacturing (PBFAM). Traditionally, depowdering is carried out manually using high-velocity air jets in a vacuum-insulated setting. Also, the part is induced with vibration forces to agitate powder particles compactly packed in tight crevices. Apart from the obvious risks in handling metal powder particles, entrapped powder particles tend to negatively affect the part quality as they interact with the as-built part during the PBFAM heat treatment post-processing stage. In summary, the effective and complete removal of entrapped powder is critical to safely handling parts and achieving superior part quality. This paper aims to automate the entrapped powder removal process using computational geometry and graph search algorithms; the objective is to find the optimum sequence of orientations that results in the complete depowdering of a part manufactured using the PBFAM process. Initially, an automatic method to detect the internal channels in a given Computer Aided Design (CAD) part file is provided. The internal channels can be detected and separated using the Boolean union and subtraction applied to the CAD body and its bounding box. The detected powder-entrapped region is converted into a triangulated mesh geometry, which is voxelized and skeletonized to form a graph data structure. The mesh skeleton, thus developed, represents the path with which the powder particles would flow along the internal channel. A ray-casting algorithm estimates the radius of the maximum inscribed sphere that would fit inside the internal channel at a particular skeletal graph location. These spheres, whose radii are embedded as node attributes to the skeletal graph, physically represent the approximate quantity of entrapped powder at that location. Subsequently, a powder flow model based on the graph structure is proposed to effectively track a powder particle’s movement inside a channel. Based on the powder flow graph model, the next best orientation to navigate the powder toward the part’s exit is identified, and the skeletal graph and its attributes are recursively modified at each intermediate orientation until all the entrapped powder particles are virtually removed from the part. Two non-trivial test cases with complex internal channels are presented to test the robustness of the proposed algorithm. The main impact of the proposed research is that this framework enables the designers to consider depowdering challenges in PBFAM during the early design stage of a part. By modifying the geometric features based on the number of orientation setups required for complete depowdering, the post-processing phase of the PBFAM process can be planned in a cost and time-efficient manner.
