Assembly sequence planning for complex real-world CAD data
Date issued
Authors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
License
Abstract
This work investigates algorithms for digital assembly sequence planning. The focus here is on finding feasible assembly sequences for a real-world data set from the automotive industry. Flexible fastening elements that have to be deformed when being disassembled, a large workspace in which the components are located and a very high number of assembled parts push conventional assembly sequence algorithms to their technical and temporal limits when it comes to real-world data. In three consecutive chapters of this work, we investigate how to push these boundaries into the feasible range, meaning that we are conducting pioneering research into the calculation of feasible assembly sequences for real-world data.
The first part of this thesis presents an approximation of a general Voronoi diagram which subdivides the workspace at the maximum clearance. We combine a common voxel propagation algorithm with a novel memory-saving data structure. We then discuss different ways to extract a general Voronoi diagram graph. This graph is a roadmap which provides a roughly estimated translational disassembly path for assembled parts, called Voronoi path. The second part of this work introduces our new path planner Expansive Voronoi Tree (EVT). For a given part, the EVT searches along the according Voronoi path for a collision-free disassembly path, consisting of arbitrary translations and rotations. Our experiments show that the EVT finds shorter paths more reliably within shorter calculation time than conventional path planning algorithms. In the third part, we present our framework for finding feasible assembly sequences. We are the first to calculate feasible assembly sequences for a real-world data set. By using two new algorithms, we detect parts that are still enclosed, which significantly reduces the number of path planning requests and, in turn, calculation time. Our framework divides the disassembly process for an assembled part into a NEAR- and FAR planning phase. Using the path planner from Hegewald et al. (2022) which is designed for deformable fasteners, we unlock parts in the NEAR range and then, in the FAR planning phase, navigate them to a nearby goal position using our EVT. The disassembly paths found are then presented in our new, informative and easy-to-read Assembly Priority Graph. Compared to other representation methods our graph is compact but also shows a high number of feasible assembly sequences. This graph is the result of the interplay of all previously mentioned findings and thus the final result of this work.