TAMPER-EVIDENT PACKAGING: A CASE STUDY
DEEP LEARNING AND MACHINE VISION FOIL INSPECTION SOLUTION
This application shows how a hybrid inspection system may be developed and deployed successfully in the field, utilizing analytical machine vision and deep learning technologies. To achieve high detection accuracy for defect features (clearly or poorly defined), the system’s strengths are effectively combined. One such system for diverse products was effectively deployed to numerous places. The technology maintained a low false failure rate while accurately detecting significant seal flaws. Finding variations in cap torque based on seal integrity and completeness was a crucial area for process optimization. The cap torque was changed to achieve tighter seals and decrease product rejects using the continuing process data.
Liquid on the cap and bottle, which could cause issues with later operations like labeling, was also discovered using thermal imaging. As a result, the overall quality improved. Additionally, the system effectively recognized even uncommon flaws, like fractured caps that would have gone unnoticed.
This program skillfully integrates cutting-edge imaging methods and analytical tools to significantly improve a crucial industrial process. The use of tried-and-true industrial thermal imaging techniques and an original combination of deep learning and discrete vision tools are essential. The outcome is a comprehensive solution with broad application in industry.
HYBRID IMAGING FOR TAMPER-EVIDENT PACKAGING
Following a malicious product contamination incident in the 1980s, the FDA mandated tamper-evident packaging for all pharmaceutical OTC products. Food and beverage manufacturers have followed that lead to help ensure a product’s quality. Capping processes impact the formation of a complete and robust seal, resulting in damaged/under-formed seals hidden under a plastic cap.
The seal is induction heated, making it adhere to the bottle rim. The foil temperature is key to the final seal quality. For this application, a thermal camera acquired an image of the foil thermal signature through the plastic cap post induction heating process. While some defects are clearly defined in the thermal images, others are more subjective and difficult to quantify. The successful inspection solution combines analytical vision tools with deep learning in a “hybrid” imaging analysis.
The system, deployed in various manufacturing locations, accurately captured important seal defects while maintaining a minimal false failure rate. Based on saved defect seal data, key areas of improvement in the process were identified, resulting in better cap torquing. This resulted in improved seal integrity and lower reject rates.