Browsing by Author "Diaz Posada, Julian Ricardo"

Now showing 1 - 1 of 1

Open Access
Optimized model-based path generation for robotic manufacturing processes
(Stuttgart : Fraunhofer Verlag, 2020) Diaz Posada, Julian Ricardo; Verl, Alexander (Prof. Dr.-Ing. Dr. h.c. mult.)
Optimized, efficient, and intuitive robotic programming is still a challenge in manufacturing and one ofthe main reasons why robots are not widely implemented in small and medium-sized enterprises and not broadly used for tasks such as milling, deburring, or welding. Reasons for this challenge are the time-consuming, complex, primarily manually optimized, and expert-dependent path generation for robotic manufacturing processes. In order to effectively and efficiently respond to the current product variability demands, small and medium-sized enterprises require easy and optimized programmable robotic manufacturing systems for achieving profitable and rapid changeover. To make up for this deficiency, this dissertation raises the following research question: How to automatically generate offline optimized paths in reaction to infeasibilities for robotic manufacturing processes under constraints and process optimization criteria? This research question is answered with the following proposed and evaluated thesis: Robotic manufacturing process paths can be automatically optimized under defined process criteria by methodologically configuring and deploying sample-based generation algorithms, based on the procedural interpretation of the product, process, and resource model-based components of the robotic manufacturing process. The basis for defining this thesis is the synergy found between four different research areas on robotics, which constitutes the state of the art of this document. These areas are: (1) the offline programming for robotic manufacturing processes, (2) the sample-based path generation, (3) the knowledge representation in robotic manufacturing and (4) the research and advances on specific robotic manufacturing processes such as welding, milling, and deburring. In this research, current advancements on these different areas are taken into account in order to develop, implement, and test the proposed architecture and methodology. The state-of-the-art analysis underscores and emphasizes that the lack of knowledge formalization, structured approach, and model-based architectures used in synergy with state-of-the-art algorithms is the reason for not having automatic optimized programming under constraints and optimization criteria for robotic manufacturing processes; resulting in the main contribution of this work. A novel approach is introduced for interpreting methodologically the components of a robotic manufacturing process using the product, process, and resource classification as input. Mathematical equations for the product features, selected by the end-user over the computer-aided design files of the workpiece in a simulation environment, are automatically generated based on models of each individual product feature. Likewise, the manufacturing process parameters are semantically described and used in order to relate the process and the resources, which are articulated industrial robots for the use-cases studied in this dissertation. The approach interprets the product, process, and resource components for simplifying the path planning problem. This is achieved by parameterizing the translational product constraints and the optimal or reference rotations of the product features in one dimension and configuring the process degrees of freedoms into a novel robotic manufacturing processes configuration space. This configuration space is introduced with the goal to relate product constraints with the degrees of freedom and constraints of the manufacturing process and to simplify the complexity of the problem by allowing its solution in reasonable computation times for industrial and practical applications as demonstrated in the implemented and evaluated use-cases. Three different use-cases are presented for simulating and evaluating the optimal path planning architecture and approach. The purpose of the evaluation in three different use-cases is to demonstrate the configurability of the approach and the re-usability of the defined models and implemented functions among different robotic manufacturing processes and criteria. Optimized robotic paths are generated for the manufacturing processes in reaction to infeasible robot configurations such as collision, unreachabilities, or maximum joint limits. The first use-case explores automatic and optimized collision avoidance in robotic welding. The second use-case explores the optimization of robot stiffness during robotic milling by using the intrinsic degree of freedom of the milling process. The third use-case demonstrates the use of a two dimensional laser scanner for sensor-guided robotic deburring which is optimized in order to get favorable measurements while assuring the deburring constraints. The simulation and validation of these three different robotic manufacturing use-cases validate the configurability of the architecture and approach. Furthermore, it is demonstrated that optimized paths for manufacturing processes can be automatically computed by using the novel approach and architecture. The approach introduced in this thesis show how the state-of-the-art development in robotic offline programming software, sample-based motion planning, robotic knowledge description, and modeling of specific robotic manufacturing processes can be used as input for improving robotic motions after further modeling the robotic manufacturing process and for generating automatic robot programs by using the proposed concept named Automatic Optimized Offline Programming (AOLP). This contribution demonstrates that the required robot programming time and the final robot process quality is improved by using the proposed approach and architecture in the implemented use-cases. Comparison between the widely used teach-in approach and a commercial offline programming software demonstrates this improvement. This approach verifies that the programming of robots does not require experts if robotic manufacturing models are used; that the robotic manufacturing processes can be automatically optimized, using sample-based algorithms after the methodological interpretation of the robotic manufacturing process; and that the general architecture can be used for several use-cases which reflect the solution to the above-defined research problem. This work concludes with an outlook on further potentials of the optimized path planning for robotic manufacturing.