Refine
Document Type
Language
- English (2)
- Multiple languages (1)
Is part of the Bibliography
- no (3)
Keywords
- 01 (1)
- AI (1)
- Assembly Sequence (1)
- Computer-aided assembly planning (1)
- High-Mix Low-Volume (1)
- Industry 4.0 (1)
- KI (1)
- Künstliche Intelligenz (1)
- LCD-Displays (1)
- Production planning (1)
Institute
Electronic appliance manufacturers are facing the challenge of frequent product orders. Based on each product order, the assembly process and workstations need to be planned. An essential part of the assembly planning is defining the assembly sequence, considering the mechanical product’s design, and handling of the product’s components. The assembly sequence determines the order of processes for each workstation, the overall layout, and thereby time and cost. Currently, the assembly sequence is decided by industrial engineers through a manual approach that is time-consuming, complex, and requires technical expertise. To reduce the industrial engineers’ manual effort, a Computer-Aided Assembly Sequence Planning (CAASP) system is proposed in this paper. It compromises the components for a comprehensive system that aims to be applied practically. The system uses Computer-Aided Design (CAD) files to derive Liaison and Interference Matrices that represent a mathematical relationship between parts. Subsequently, an adapted Ant Colony Optimization Algorithm generates an optimized assembly sequence based on these relationships. Through a web browser-based application, the user can upload files and interact with the system. The system is conceptualized and validated using the CAD file of an electric motor example product. The results are discussed, and future work is outlined.
CSOT (China Star Optoelectronics Technology), auch Shenzhen Huaxing Photoelectric Technology genannt, ist der führende LED-Lieferant der TCL Group und der zweitgrößte Produzent von LCD-Displays weltweit. Das Unternehmen nutzt Künstliche Intelligenz (KI) zur Automatisierung von Fehlerprüfungsprozessen, um seinen Wettbewerbsvorteil zu erhalten und auszubauen. Das Labeln (dt. Beschriftung/Etikettierung) von Datensätzen für das Training mit aktuellen KI-basierten Methoden ist jedoch zeit- und arbeitsintensiv und erfordert insgesamt bis zu 1 500 Stunden Trainingszeit für eine typische Produktionsfabrik mit mehreren Produktionslinien.
Um dieser Herausforderung zu begegnen, entwickelt das Hong Kong Industrial Artificial Intelligence and Robotics Centre (FLAIR) – gemeinsam initiiert vom RWTH Aachen Campus und dem Hong Kong Productivity Council – neue KI-basierte Segmentierungs- und Klassifizierungstechniken. Mit diesen Ansätzen wird das Data-Labeling von 20 Sekunden auf weniger als eine Sekunde reduziert. So werden mehr als 1 400 Arbeitsstunden pro Fabrik eingespart, was großes Potenzial für Qualitätsverbesserungen im Produktionsmanagement und geschätzte Einsparungen von mehr als 84.000 Euro bedeutet.
The operation of CNC milling is expensive because of the cost-intensive use of cutting tools. The wear and tear of CNC tools influence the tool lifetime. Today’s machines are not capable of accurately estimating the tool abrasion during the machining process. Therefore, manufacturers rely on reactive maintenance, a tool
change after breakage, or a preventive maintenance approach, a tool change according to predefined tool specifications. In either case, maintenance costs are high due to a loss of machine utilization or premature tool change. To find the optimal point of tool change, it is necessary to monitor CNC process parameters during machining and use advanced data analytics to predict the tool abrasion. However, data science expertise is limited in small-medium sized manufacturing companies. The long operating life of machines often does not justify investments in new machines before the end of operating life. The publication describes a cost-efficient approach to upgrade legacy CNC machines with a Tool Wear Prediction Upgrade Kit. A practical solution is presented with a holistic hardware/software setup, including edge device, and multiple sensors. The prediction of tool wear is based on machine learning. The user interface visualizes the machine condition for the maintenance personnel in the shop floor. The approach is conceptualized and discussed based on industry requirements. Future work is outlined.