After discussing in the previous parts about regulations, let's analyze more in detail the systems and technologies available on the market today to achieve 100% visual inspection.
There are three main basic kinds of visual inspection systems:
· Manual inspection
· Semi-automatic inspection machines
· Fully automatic inspection machines
The very first inspection system for parenteral products was introduced around 1950 and was naturally based on human inspection (MVI Manual Visual Inspection).
Since then there has been a constant evolution and several improvements in all involved procedures such as inspectors training, working conditions, etc.
A key factor to be underlined is that inspection performed by MVI is still considered as the quality level reference, especially when comparing and validating new automatic systems to be put in production.
However, it’s also very important to underline how manual inspection is
able to achieve reference good and stable performances only under optimal conditions such as:
? Periodic breaks to avoid tiredness of inspectors (at least 5’ rest every hour but it is recommended 20’ inspection followed by 5’ rest for a total duration not longer than 4h )
? Each inspector must have been trained and tested properly. Periodic check and re-qualification of inspectors is also recommended at least every year
? Low inspection speed (recommended 5” against white and 5” against black background per product equivalent to about 300-350 containers per hour)
? Optimal lighting conditions (recommended 2000-3750 lux diffused light against black and white background by USP and EP) and a Color Reproduction Index CRI > 80% (recommended >90%).
? Comfortable working conditions (temperature not higher than 25 C°, relative humidity and air speed controlled, noise level below 55 dB, comfortable seating)
? Proper handling of containers by hands or transport system for semiautomatic inspection
? Containers not bigger than 100ml in clear glass, transparent and clear liquid (inspection performances decrease quickly for bigger container’s size and lower transparency)
Figure 1: Standard bench for manual inspection
Whenever anyone or more of above condition is/are not met, the overall performances of MVI will be reduced. It is therefore required a constant work and effort to maintain stable results and performances from MVI including the management cost of several if not many inspectors required to achieve an inspection speed high enough to match production speed.
When considering the total cost required by MVI it is important to account all sub-costs related to all different activities and facilities involved: inspection workstations, management of working and resting turns, the distribution&recollection of products before/after inspection, the procedures to make sure good and bad products are properly handled and never mixed or confused, the counting and reconciliation of batch statistics.
Semi-automatic inspection is basically manual inspection assisted in handling.
One automatic machine is used to handle the containers movement, spinning and sometimes on good/reject selection to reduce the operator manual workload and allow him to completely concentrate on the inspection task and be less tired and distracted.
There is no real separation between manual and semi-automatic when we consider visual inspection performances.
The inspection system is exactly the same (human eye + brain) therefore the differences in actual performances will be only related to the better comfort and less manual work the operator is loaded with when using semi-automatic inspection systems.
Automatic Inspection Machines
First automatic inspection machines (AVI Automatic Visual Inspection) have been developed and introduced since 1980’s.
Compared to MVI and semi-automatic inspection machines, AVI were for the first time able to judge without any human assistance each container's quality level and decide and manage it's acceptance or rejection from production stream.
Figure 2: One of first inspection machines introduced in 1980's
First AVI were initially based on simple light sensors and able to check only for particles (foreign moving bodies into liquids) and filling level according to changes in light transmission. Their performances were severely limited by the very basic technologies available at that time. Those inspection systems weren't really "able to understand" the inspection process and what was visible inside containers.
Soon afterwards, several key technologies started undergoing an impressive development in performances and reduction of costs and dimensions which lasted for more than 20 years and finally allowed automatic inspection machines to mimic much more closely the human eye+brain system and to begin actually "watching" and "understanding" the products.
Figure 3: Exponential development of CPUs
The main areas responsible for this improvements were computers, software and cameras, but many others contributed as well such as lighting and motion systems, PLC and automation controllers, software algorithms and development tools.
Hence, since 90's the basic light-change detection systems have been gradually abandoned and today almost all automatic inspection machines are based on the camera + software systems to analyze and evaluate the images of containers.
This more sophisticated approach allows much better analysis of the images and can improve inspection performances and reduce false rejects.
Figure 4: Flow of information in Manual and Automatic systems
Particles Detection in Automatic Inspection Systems
The detection of foreign particles is based on the same basic principles of manual inspection:
? liquid is set in movement by spinning or swirling of the container
? stop and check the container, take several images of liquid content
? analyze images to find moving particles dragged by moving liquid
? accept or reject the inspected container according to specified criteria about brightness, number and size of particles
The key point here is movement: in both inspection systems the movement of particles along with spinning liquid is exploited to increase the sensitivity to even smaller particles
Lighting and Imaging of Containers
Even in automatic inspection it is common to execute a double inspection against black and white backgrounds in order to maximize probability of detection of all kinds of particles. Some kinds of particles are in fact easier to spot in one or another condition according to how they reflect and adsorb the light.
The difference between manual and automatic inspection is that in the latter one, the source of light is engineered to exploit those features of particles to the maximum extent.
Backlight, which is equivalent to manual inspection against a white background, is designed to exploit the adsorption of light by the particles which will appear as moving dark spots against a white background.
It is more efficient in the detection of opaque particles, very dark and generally reflecting very little light such as carbon particles, burn residues, some hairs, parts of insects, some kind of rubber particles.
Figure 5: Example of vial's backlight image
Bottomlight, equivalent to manual inspection against a black background, is based on the reflection of light by the particles (the Tyndall effect). This allows even very small particles to become visible to the eye or the camera, but the system is effective only with particles able to reflect a significant amount of light and is very sensitive to small air bubbles, which also are very efficient in light reflection.
Figure 6: Example of vial's bottomlight image
Overall, a good inspection system needs to be well balanced in all its aspects: lighting, spinning, imaging and processing in order to achieve a good balance and performances.
For example a very sensitive system can be very effective in detecting and rejecting even smallest particles, but at the same time it will detect and reject also air bubbles, reflections of light or very small particles therefore increasing too much the overall and false reject rate.
Other Controls in Automatic Visual Inspection Systems
In early automatic inspection machines only moving particles and filling level controls were implemented. Later the development of components and technologies has led to the introduction of several additional quality controls executed automatically through camera images:
? Cosmetic/sidewall inspection: for the detection of cracks, dirt and other defects of the container.
? Inspection of freeze dried cakes
? Sealing/capping inspection: for the detection of defects in stoppering, alu-sealing, crimping and closing of containers.
? Shape and dimensions of the containers
? Color and transparency of the solution
Figure 7: A modern Fully Automatic Inspection Machine
END OF PART 6