Improving Additive Manufacturing production planning: A sub-second pixel-based packing algorithm

Abstract

Additive Manufacturing (AM), the technology of rapid prototyping directly from 3D digital models, has made a significant impact on both academia and industry. When facing the growing demand of AM services, AM production planning (AMPP) plays a vital role in reducing makespan and costs for AM service companies. This research focuses on the AMPP problem under the unrelated machine environment and two-dimensional irregular packing constraints such that the makespan can be minimized. Since there is no promising sub-second algorithm for solving the 2D irregular packing sub-problems in current literature, the efficiency of checking the packing feasibility of batches is the bottleneck of algorithms of AMPP. Therefore, we propose an efficient pixel-based AM packing algorithm (PAMPA) which can tackle irregular packing sub-problems that allow hole filling and free rotation. In PAMPA, parts are placed one by one in a Bottom-Left (BL) way and the rotation angle is determined by a novel angle selection heuristic. A new codification method and contour-guided method are proposed to accelerate separating overlap. The placement sequence is optimized by a biased random-key genetic algorithm (BRKGA). Furthermore, by adopting PAMPA, we introduce a variable neighbourhood search (VNS) framework to solve the AMPP. Finally, new instances are generated to conduct experiments. Based on the new instances and instances from (ESICUP), the efficiency of PAMPA is analysed and verified. The proposed VNS framework can reduce the makespan by 8 h on average compared with the initial solution, which illustrates that our PAMPA is beneficial for AMPP. Some interesting insights are also revealed and discussed based on the case study.