Model- and data-driven process and quality optimization in container glass production
Glass containers are produced through a complex multi-stage process, which converts raw materials into molten glass, enables gob forming and feeding and forms gobs into glass containers. At the end of the process, quality inspection machines shall examine specific quality characteristics of the glass containers, such as wall thickness. Containers that do not qualify to the specified grade are thus rejected. On the one hand, interactions among process variables and wall thickness are complex. Potential influencing factors of the wall thickness are therefore still not completely clear. On the other hand, structures and format of the data from different measurement systems are inhomogeneous, which makes identification and merging of data of single glass containers extremely difficult. According to the findings of IPGR, deviation of the wall thickness distribution in the current glass container production still shows potential for improvement.
Therefore, PLT and IPGR initiate a long-term research plan, which aims at an overall optimization of the production process, plant and product through developing concepts and prototypical tools for contextualization, integration, preprocessing and analysis of process data. As pioneers in the digitalization trend, IPGR is working together with its members from the container glass industry in this area, focusing on research and development in the field of data-driven process and quality optimization.
The previous and future research projects, which are part of the long-term research plan, focus on the following aspects:
- Investigation of the relationship between the process parameters and process variables and the wall thickness ratio using methods of the Knowledge Discovery in Databases
- Development of software tools for accessing, pre-processing and visualizing the data from production process
- Development of methods for representing and analyzing interacting relations between process parameters and product quality values, using interdisciplinary theories e.g. graph theory, information modelling, mathematical modelling etc.
- Development of concept for integration of data generated by machines and sensors
- Prediction of product quality by deploying AI-based methods