Entwicklung von kamerabasierten Prüfsystemen mit künstlicher Intelligenz für Silikon-Spritzgussteile

  • Johannes Reinwein

Student thesis: Master's Thesis

Abstract

In the recent years, there has been a steady increase in the level of quality requirements from the customers of Starlim Spritzguss GmbH. This leads to a particular challenge due to the high production volumes and low unit costs associated with the production of silicone injection moulded parts. Manual, optical quality control is often time-consuming, error-prone, and labour-intensive. Conventional camera-based inspection systems are typically unable to reliably detect all possible types of defects, and the inspection process must be customised for each product. However, technological progress in the field of artificial intelligence (AI) is opening new opportunities for developing automated inspection systems that could fulfil the requirements of comprehensive quality control while simultaneously reducing implementation costs.
The aim of this master thesis is to train various models with artificial intelligence for the quality control of silicone injection moulded parts and to test them in experimental studies. At the beginning, typical defect types and application areas of relevant components are analysed. For better understanding of the topic, a theoretical review of common image processing techniques is following, along with an introduction to the field of artificial intelligence. The focus is on functionality, training and evaluation of models with neural networks, which are frequently used in systems with artificial intelligence. Based on these findings suitable hardware and software will be selected for the experimental investigations. These investigations aim to demonstrate the advantages and disadvantages of AI-supported quality control under production conditions. Finally, the applicability of the investigated methods for future quality control in production will be evaluated.

"Make visible what, without you, might perhaps never have been seen."
Robert Bresson
Date of AwardNov 2024
Original languageGerman (Austria)
SupervisorGerald Zauner (Supervisor)

Cite this

'