Smart quality control Detecting injection moulding defects more cost-effectively with AI and robots

Related Vendors

TH Köln

HASCO Hasenclever GmbH + Co KG

BORIDE Engineered Abrasives

Image analyses to detect defects in injection-moulded plastic parts used to be expensive and time-consuming. Cologne researchers want to change this using machine learning (ML) ...

Injection moulding should result in plastic products with visually sophisticated surfaces — especially when it comes to the automotive industry and its requirements. Image analysis methods can be used to detect faulty parts, as researchers at TH Köln point out. However, these methods must first be trained in a time-consuming and therefore expensive process. The reason for this is that the defect patterns in plastic parts are very diverse, as the report continues. Therefore, the applicability of artificial intelligence in the production of injection moulded parts is limited. In order to improve and accelerate quality control, TH Köln has now developed and tested an automated, camera-based AI option with two industrial partners.

Does the injection moulded part fit or not?

In plastics processing, the rapid detection of rejects is of great interest. Especially for components that are produced in large quantities, for which injection moulding is actually ideal. Particularly in the context of quality control, it is very time-consuming to collect and verify the necessary training data during operation. Despite the complexity of the error patterns, the use of AI still makes sense because manual quality control takes a long time and does not always lead to the desired result. The aim of the project “QuKu-ML: Camera-based quality assessment in plastic injection moulding using ML strategies” was to simplify the quality control of a component produced by an injection moulding machine for the automotive industry using an algorithm.