The applications of artificial intelligence (AI) are increasingly becoming tangible and are profoundly affecting aspects of our daily lives. Can the potential of neural networks be applied to quality control for pharmaceutical products? What prospects does AI have in this sector? These are some of the questions Davide Luisari, Design Manager for Visual Inspection, Bonfiglioli Engineering, answered for us.
Visual inspection inspects the integrity of pharmaceutical packaging, identifying the presence of cosmetic and particulate defects. Products that do not comply with the expected quality standards are discarded.
Visual inspection uses a system of very high-resolution cameras that collect several images of each product to be analyzed, photographing all the sample’s various components. These images are then inspected for irregularities (such as scratches, inclusions of foreign bodies, or dents) that could damage packaging integrity and compromise the contents.
The images can be analyzed by specialized operators or by a computer equipped with a data processing algorithm. Automated control provides decidedly better results in terms of testing speed, cost, accuracy, repeatability, and reliability.
All these advantages are further amplified when artificial intelligence is added to the capabilities of automated control.
Automated control uses algorithms to analyze the images collected by the cameras and determine whether a product is compliant. While this method is both efficient and accurate, AI has a unique feature that makes it more competitive: machine learning. AI learns from its mistakes and corrects itself, becoming increasingly reliable. Algorithms, in fact, provide their results based on parameters that are established during machine set-up. This means that, in unusual situations that don’t match the expected case studies, the algorithms are unable to analyze the problem and simply reject the product, considering it non-compliant since it does not fall within the predefined parameter set.
AI on the other hand, works from a set of images that are catalogued as compliant or non-compliant. It can determine whether the product it is analyzing belongs to either category by verifying many parameters that it checks in sequence. Artificial intelligence does not use a rigid algorithm but self-calibrates the parameters it uses to provide the result, with no need for operator intervention. AI “learns” and can provide correct results even in unexpected situations.
In many fields, human capabilities still cannot be defeated. There is a heated debate on the possibility of using neural networks to provide services in complex areas such as education, care for the elderly and healthcare. Although a variety of experiments are currently underway, these applications are still very far from becoming reality. In other areas however, AI has already proved to be far faster and more accurate and reliable than humans in performing specific tasks.
For example, the human eye and mind can be deceived by optical illusions that distort how we interpret what we see. A machine would not make such mistakes and can provide reliable, objective, and repeatable results even in cases where a human operator might have doubts or uncertainties. The human eye also has anatomical limitations that make it difficult for an operator, even a skilled one, to detect irregularities measuring 100 microns in diameter. Bonfiglioli Engineering automatic visual inspection machines can detect much smaller particles, making them much more sensitive and accurate.
While artificial intelligence is not “smarter” than humans in absolute terms, it can certainly perform visual inspection more quickly, reliably, and objectively.
Automatic visual inspection is less expensive than you might think. It is truly advantageous in terms of cost, since using this type of technology allows machines to operate at speeds higher than those possible for traditional controls. Automated system guarantees higher speed performance, with constant accuracy and repeatability. This enhanced speed always results in significant economic returns, as well as better quality.
Bonfiglioli Engineering provides Automatic Visual Inspection (AVI) solutions with Artificial Intelligence (AI). Each solution is unique and can be custom built to suit your needs. The system can be combined with other testing methods too, to provide a single solution for multiple testing.
Want to know more? Connect with our experts for consultation.