Product novelty 16. September 2024
AI in the plastics industry: quality prediction for injection moulded parts
Initial situation
Today, injection moulding companies mainly ensure part quality through statistical checks. However, this process is largely manual and therefore time-consuming and error prone. Fully automated quality determination in combination with artificial intelligence (AI) offers new possibilities. By analysing process data directly after the injection moulding cycle, critical changes in the process can be identified and the component quality predicted.
Solution
The use of AI makes it possible to predict the component quality directly after the injection moulding cycle based on the process data. This is realised through initial training in which the recorded component qualities are correlated with the process data. All this data can be assigned to a component using a serial number, in this case a lasered DataMatrix code. An important quality feature, which also reflects the process control, is the shrinkage of the plastic components. After production, these components must be stored for 72 hours and then re-measured. A mobile logistics robot is used at the IWK for this purpose, which transports the containers with the components from the system to the measuring cell and the warehouse. If an anomaly is detected in the process or component, these are checked again in the measuring cell in order to retrain the AI model and successively optimise it.
Further developments
The measuring cell from Kistler and the prediction model are being further developed together with the ILT Institute for Lab Automation and Mechatronics at OST and the IWK Institute for Materials Technology and Plastics Processing and tested on practical components and industrial environments with the help of partners from industry, medmix AG and Weidmann Medical Technology AG (Fig. 1). The aim is to develop a multifunctional and autonomous measuring cell for predicting component quality that can be used for both random sampling and permanent monitoring.
Today, injection moulding companies mainly ensure part quality through statistical checks. However, this process is largely manual and therefore time-consuming and error prone. Fully automated quality determination in combination with artificial intelligence (AI) offers new possibilities. By analysing process data directly after the injection moulding cycle, critical changes in the process can be identified and the component quality predicted.
Solution
The use of AI makes it possible to predict the component quality directly after the injection moulding cycle based on the process data. This is realised through initial training in which the recorded component qualities are correlated with the process data. All this data can be assigned to a component using a serial number, in this case a lasered DataMatrix code. An important quality feature, which also reflects the process control, is the shrinkage of the plastic components. After production, these components must be stored for 72 hours and then re-measured. A mobile logistics robot is used at the IWK for this purpose, which transports the containers with the components from the system to the measuring cell and the warehouse. If an anomaly is detected in the process or component, these are checked again in the measuring cell in order to retrain the AI model and successively optimise it.
Further developments
The measuring cell from Kistler and the prediction model are being further developed together with the ILT Institute for Lab Automation and Mechatronics at OST and the IWK Institute for Materials Technology and Plastics Processing and tested on practical components and industrial environments with the help of partners from industry, medmix AG and Weidmann Medical Technology AG (Fig. 1). The aim is to develop a multifunctional and autonomous measuring cell for predicting component quality that can be used for both random sampling and permanent monitoring.
OST Ostschweizer Fachhochschule
IWK Institut für Werkstofftechnik
und Kunststoffverarbeitung
Contact
Mr. Jasper Hollender