Customers are demanding ever-more sophisticated inspection performance to ensure constant maximization of output quality and allow real-time resolution of any potential deviation from production standards. Computer Vision by SACMI, to be presented K 2019 in Düsseldorf at Stand A63 in Hall 13, is an application that executes on-line preform quality control using polarized light.
Generally used to identify abnormal stress on preform surfaces, polarized light inspection previously had the drawback of being impossible to perform directly on-line because traditional algorithms were not up to the task. SACMI succeeded in replacing traditional sampling checks – performed manually on test benches – with a fully automated control system that works in combination with advanced AI algorithms.
The company will display PVS10L, equipped with the new, patented SACMI control system. With the PVS10L, the preform line is transformed into a system that can simply and quickly ‘self-learn’ all the required checks. This simplifies QC procedures and improves working conditions for operators, who can work directly on the line without having to manually set inspection recipes.
SACMI’s computer vision range depends on the drive to develop algorithms reliable enough to automate such tasks. This approach is becoming the standard in various fields and is leading to the development of new applications and other products and stages of the production process. The ultimate goal is to boost quality control, extending it from inspection of every individual product to govern the entire production process, thus maximizing line efficiency by objectively identifying – without variables that depend on human intervention – the problem type and its origin.
Development of AI algorithms – used together with traditional video camera inspection techniques – is also being driven by the need to simplify the system. This results in a dual advantage – the possibility of fielding new checks and, at the same time, simplifying the checks themselves with systems that can self-learn and thus respond to the most complex inspection needs of the plastics and metal graphics industries while minimizing waste and optimizing process costs.