CPS-based manufacturing workcell for the production of hybrid medical devices

• Published in: Journal of ambient intelligence and humanized computing Vol. 12; no. 12; pp. 10865 - 10879
• Main Authors: Ho, Nicholas, Wong, Pooi-Mun, Hoang, Ngoc-Son, Koh, Dun-Kai, Chua, Matthew Chin Heng, Chui, Chee-Kong
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2021; Springer Nature B.V
• Subjects: Anthropocentrism; Artificial Intelligence; Artificial neural networks; Augmented reality; Cell growth; Computational Intelligence; Computer architecture; Cost reduction; Cyber-physical systems; Deep learning; Design; Digital technology; Engineering; Hybrid systems; Image processing; Internet of Things; Machine learning; Machine vision; Manufacturing; Medical device industry; Medical devices; Medical electronics; Medical equipment; Medical imaging; Medical technology; Neural networks; Original Research; Preventive maintenance; Prototypes; Real time; Robotics and Automation; Subsystems; User Interfaces and Human Computer Interaction; Intelligent manufacturing workcell; Hybrid medical device; Deep learning machine vision; Cyber-physical system; Internet-of-things; Augmented reality
• ISSN: 1868-5137; 1868-5145
• DOI: 10.1007/s12652-020-02798-y

Cyber-physical system (CPS) can increase efficiency and lower costs in manufacturing operations. CPS and related technologies can help medical device manufacturers to meet the rising demands for quality and affordable healthcare products and services. In this paper, a novel architecture for CPS-based manufacturing workcell systems was proposed. Its contribution is to provide quality inspection and operator guidance in manual assembly systems, specifically for the production of hybrid medical devices. The proposed system involves the integration of enabling technologies including Machine Vision (MV), Internet-of-things (IoT) and Augmented Reality. The group illustrated and evaluated the implementation of two CPS prototypes based on the proposed architecture on the fabrication of a specific hybrid medical device, verifying their effectiveness as guidance tools for operators. The two prototypes, CPS-1 and CPS-2 differ in their MV subsystems by having non-deep learning and deep learning-based image processing methods respectively. Experimental results illustrate that, when exposed to various changing conditions, the Regional Convolutional Neural Network (R-CNN) of CPS-2, on average, has outperformed the non-deep learning image processing method of CPS-1 by 42.3% and 8.5%, in terms of accuracy and precision respectively. Hence, this presents evidence on the usefulness of R-CNN-based MV system to increase the flexibility and adaptability of the CPS in terms of object monitoring and reporting.