Guide to defects detected by Visual Inspection
Visual Inspection technology uses an image detection system and a set of algorithms to interpret data and identify conditions of non-compliance.
Visual Inspection technology can detect both cosmetic defects and the particle-size contamination of its contents in product packaging. Inspections carried out by the machine can be summarized as follows:
Cosmetic inspection where defects referred to the external package and/or defects referred to components of the container itself are investigated. The following irregularities are searched for:
- Rubber Stopper Absence
- Non-Conforming Rubber Stopper Shape
- Flip-Off Absence
- Non-Conforming Flip-Off Shape
- Non-Conforming Flip-Off Color
- Damaged Flip-Off
- Non-Conforming Crimp
- Dented Aluminium Ring
- Scratched Aluminium Ring
- Non-Conforming Tip Shape or Ring
- Non-Conforming Fill Level
- Non-Conforming Content Color
- Non-Conforming Container Color
- Cosmetic Defect on Neck (cracks, scratches, burns, spots, shape defects)
- Cosmetic Defect on Body (cracks, scratches, burns, spots, shape defects)
- Cosmetic Defect on Bottom (cracks, scratches, burns, spots, shape defects)
Particle inspection where contaminated bodies/content inside the product are investigated in detail. The following irregularities are searched for:
- Outside particle (Cake)
- Cake volume / color / form defects
- Outside particle (powder)
- Particle contamination, all types, and possibilities (Liquid)
Machine supervised learning models are currently being applied to automated visual inspection (AVI) for difficult to inspect products such as BFS (Blow-Fill-Seal) containers. Artificiall intelligence (AI) based programs have shown remarkable success in automating tasks and improving the decision-making process. They have proven highly effective in enhancing image classification accuracy and identifying product variations (such as forming defects and bubbles) that simple silhouette inspections cannot detect. AI neural networks (NNs) trained on customers’ containers learn to distinguish between conforming and defective products, even when certain container features might resemble defects. Different types of NNs, each specialized for different tasks, combined with traditional tracking algorithms, enable effective AVI of containers to achieve an acceptable false rejection rates (FRR).
Citations
This article is inspired by information from Bonfiglioli Engineering.
- Bonfiglioli Engineering. “Guide to defects detected by Visual Inspection” – Bonfiglioli Engineering, BONFIGLIOLI ENGINEERING S.r.l