Non-contact techniques for the quality analysis of PET bottles

This work was motivated from finding a complementary way of tuning and controlling the machine parameters of Injection Stretch Blow Molding process. In the current approach, a specialized technician detects the bottle defects by visual inspection and corrects the machine parameters using its own experience or indications obtained by previous statistical analyses. As all human based operations, inherent limitations are that the results are influenced by the operator skills; in addition, the experience can be hardly converted into a database, which could be used for the process optimization. The solution investigated in this work is to replace visual inspection with an image processing system. A prototype for offline analyses of PET bottles was designed in order to have a resolution allowing to identify the most common bottle defects. The acquired images were analyzed with algorithms implemented in LabVIEW. Results showed that this system can off-center gate, haze and pearlescence with a repeatability and reproducibility sufficient for the identification of bottles with manufacturing defects.