TY - CHAP A1 - Böhm, Stefan-Andreas T1 - AI Approaches to Optimize Human-Machine Collaboration in Manufacturing Facilities with IoT-Ready Machinery T2 - IoT '20 Companion: 10th International Conference on the Internet of Things Companion, October 2020 N2 - Modern production facilities are becoming increasingly complex, as companies battle to meet the increasingly customer-specific demands with shorter lead times and higher efficiency. For the employees in the factory, who often bear the brunt of making short-term process decisions, this complexity is becoming unmanageable. The PhD project presented in the following aims to decrease the complexity of modern manufactory facilities by means of artificial intelligence. The focus is on product data analysis and manufactory process planning. Different approaches of artificial intelligence for use in a Self-Adapting Smart System (SASS) are to be investigated. KW - Self-Adapting Smart System Y1 - 2020 UR - https://doi.org/10.1145/3423423.3423471 N1 - Article No.: 23 SP - 1 EP - 5 ER - TY - CHAP A1 - Kortüm, Christian A1 - Böhm, Stefan-Andreas ED - Schajer, Gary T1 - Holistic approach for geometric and technological similarity search in the production for profiled elements in the wood working industry for a more efficient production T2 - Proceedings of the 25th International Wood Machining Seminar N2 - In the whole field of woodworking, from carpentry to industry, new products are created and designed every day and manufacturers wonder if similar workpieces have already been produced in the company and if corresponding drawings, information, tools, jigs, etc. can be reused. Under certain circumstances, considerable duplication of work can be avoided and a more efficient production is possible. In order to find similar workpieces, it must be clarified what exactly is meant by similarity. The different fields of knowledge do not provide a clear definition. However, similarities are always used to compare objects. In woodworking, objects can be a wide variety of components and products, such as carcase sides, furniture fronts, fittings, tools or profile strips. The search for similar profiled elements – related to the potential, data structure and algorithms – is the focus of this paper. There is a wide range of different profiled elements, which are used constructively as well as decoratively, e.g. for baseboards, door frames, window frames or decorative moulding profiles. These profiles are manufactured in a multi-stage production process in which the workpieces are first machined with profiled tools before surface finishing takes place. Along this production chain, tools, machine settings and, in some cases, fixtures are used that are adapted to the respective profile and can be reused for the same or similar jobs. Inheritance of information offers the greatest potential of similarity search. Inheritance means that existing knowledge in the form of drawings, CNC programs, process parameters and setup specifications, etc., is transferred from existing profiles to a new profile. By inheriting existing information, for example, the effort required for process setting can be reduced by adopting spindle positions, positions of guides and stops, or feed and cutting speeds. However, this can only be done if the process knowledge and other technological information concerning a profile are sufficiently documented. This is the only way how all information is made available for the similarity search and inheritance, and also for setting the processes when the profile is manufactured once again. The profile data must therefore be stored in a meaningful structure without loss of information. Therefore, a holistic product data model was developed that combines the geometrical with the technological profile data. The developed model allows a combined search for geometrically and e.g. materially similar products – also with a weighted evaluation of the individual criteria. The research has shown that data modelling is the essential foundation for the implementation of modern AI algorithms to enable similarity search. In conclusion, the use of a search system for geometrically and technologically similar profiles represents a real innovation for wood processing companies. This paper presents a holistic approach for geometrical and technological similarity search presenting different algorithms KW - similarity search KW - profiled elements KW - product data model KW - multi-stage production process Y1 - 2023 ER - TY - CHAP A1 - Böhm, Stefan-Andreas A1 - Kortüm, Christian ED - Schajer, Gary T1 - Machining Feature Recognition in Furniture Parts based on a Graph Neural Network T2 - Proceedings of the 25th International Wood Machining Seminar N2 - Due to the increasing customer demands, companies face the challenge of adapting their manufacturing facilities to produce batch size one parts. In this context, each component comes with specific customer requirements, significantly increasing the complexity of its manufacturing process. Traditional process planning done by employees is no longer feasible in the face of the growing complexity and diversity of batch size one parts. Therefore, there is a need for novel computer-aided process planning methods capable of autonomously learning and identifying the necessary machining features of these parts. We propose a configurable automated feature recognition framework tailored to batch size one parts to address this need. This framework comprises two main components. First, a customizable data generator is developed to create a dataset of computer-aided design models incorporating user-specific machining features. Second, a graph neural network is employed to learn these user-specific machining features by leveraging the graph structure represented by the vertices and edges in each computer-aided design model. We evaluate the effectiveness of our framework using a real-world example from the furniture industry. Specifically, we construct a dataset comprising over 7000 computer-aided design models of wooden boards using our framework's data generator. Each wooden board contains up to 30 randomly placed machining features for the application of connectors and fittings, randomly selected from 14 standard machining features commonly used in the furniture industry. An optimization algorithm adapts a graph neural network for classifying these specific machining features. The results demonstrate that our framework addresses the automated feature recognition problem without requiring data conversions, such as pixels or voxels, as other approaches often do. By introducing this configurable automated feature recognition framework, we solve the challenges posed by batch size one manufacturing. The framework's effectiveness and accuracy are demonstrated through extensive testing on the example of the furniture industry, showcasing its potential for enhancing manufacturing processes in various domains KW - Furniture Industry KW - Automatic Feature Recognition, KW - Machining Features KW - Machine Learning, Graph Neural Networks Y1 - 2023 ER - TY - INPR A1 - Böhm, Stefan-Andreas A1 - Neumayer, Martin A1 - Kramer, Oliver A1 - Schiendorfer, Alexander A1 - Knoll, Alois T1 - Comparing Heuristics, Constraint Optimization, and Reinforcement Learning for an Industrial 2D Packing Problem N2 - Cutting and Packing problems are occurring in different industries with a direct impact on the revenue of businesses. Generally, the goal in Cutting and Packing is to assign a set of smaller objects to a set of larger objects. To solve Cutting and Packing problems, practitioners can resort to heuristic and exact methodologies. Lately, machine learning is increasingly used for solving such problems. This paper considers a 2D packing problem from the furniture industry, where a set of wooden workpieces must be assigned to different modules of a trolley in the most space-saving way. We present an experimental setup to compare heuristics, constraint optimization, and deep reinforcement learning for the given problem. The used methodologies and their results get collated in terms of their solution quality and runtime. In the given use case a greedy heuristic produces optimal results and outperforms the other approaches in terms of runtime. Constraint optimization also produces optimal results but requires more time to perform. The deep reinforcement learning approach did not always produce optimal or even feasible solutions. While we assume this could be remedied with more training, considering the good results with the heuristic, deep reinforcement learning seems to be a bad fit for the given use case. KW - Computer Science - Artificial Intelligence Y1 - 2021 U6 - https://doi.org/10.48550/arXiv.2110.14535 ER - TY - INPR A1 - Böhm, Stefan-Andreas A1 - Zagar, Bare Luka A1 - Riß, Fabian A1 - Kortüm, Christian A1 - Knoll, Alois Christian T1 - Automated Feature Recognition in Surface Cad Models Based on Graph Neural Networks N2 - Driven by increasing customer demands, manufacturing processes encompass increasingly intricate workflows. The industry relies on computer-aided process planning software to effectively manage the intricacies of these complex manufacturing processes. This software is crucial to analyze computer-aided design data for a product and determining the required machining steps. However, a notable challenge arises, particularly in the case of custom products, where the machining steps can significantly vary depending on the available machinery and the employees' preferences.This study introduces a configurable automated feature recognition framework based on expert knowledge. Experts can encode their insights and expertise within this framework using a configurable synthetic data generator. Modern graph neural networks learn from the data generated by this data generator, achieving an average recognition accuracy up to 80% (F1-score) with a runtime performance of 21.98 milliseconds per model. Importantly, it accomplishes this even when confronted with highly intersecting machining features without requiring data conversion into alternative formats, such as voxel or pixel representations. KW - 3D deep learning KW - CAD KW - Graph neural networks KW - Machining feature recognition KW - Snthetic data generation KW - Surface models Y1 - 2024 U6 - https://doi.org/10.2139/ssrn.4772779 ER - TY - CHAP A1 - Böhm, Stefan-Andreas ED - Tomforde, Sven ED - Krupitzer, Christian T1 - Adaptable Machining Process Identification based on Expert Knowledge and Artificial Intelligence T2 - Organic Computing Doctoral Dissertation Colloquium 2021 N2 - The demand for product individuality increased enormously in recent years and thus affects directly manufacturers and their employees. Due to the increasing demand for batch-size-one production, every product needs specific manufacturing processes. Usually, employees determine these manufacturing steps with provided product data. This research proposal aims to contribute to the extraction of machining processes from product data and their assignment to suitable machinery. We plan to develop an organic computing system based on artificial intelligence methods to solve these problems by including customer-specific designs and employee expertise. The overall objective is to support employees in manufacturing facilities by simplifying the manufacturer and customer interaction. KW - organic computing KW - artificial intelligence, KW - reinforcement learning KW - internet of things KW - manufacturing systems Y1 - 2022 U6 - https://doi.org/10.17170/kobra-202202215780 SP - 1 EP - 14 PB - kassel university press CY - Kassel ER -