Inspection is being revolutionized through new ways of managing it. Put differently, a vision system doesn’t just need to be accurate, precise and useful from a feedback on process and predictive maintenance perspective. A vision system must (or should) also be easy to use, even for an unskilled operator.
With SACMI, inspection has now evolved into Classy AI, a tool that lets you manage the system via a user-friendly dashboard and immediately classify defects by type, from the most commonplace to those that can only be interpreted by highly qualified technicians.
And as complexity increases - in terms of the technology, the huge variety of sizes and raw materials and the quantity and nature of the defects that need intercepting - system management therefore needs to become simpler.
At the K fair in Düsseldorf, from 19 to 26 October 2022, SACMI will present a Classy AI that’s been integrated in the new PVS156 preform vision system, in turn installed on their IPS 300 preform press.
The Simplicity of Complexity. Thanks to Artificial Intelligence. Classy AI is a revolution SACMI believes in because it takes the company exactly where their customers want to go: towards more accurate control and, above all, towards simpler operator workflows that yield immediate, tangible advantages.
To illustrate more clearly, let's look at the current state of the art. For years now, SACMI has been working to make the operator interfaces on their systems as user-friendly as possible. In parallel, the technological quality and complexity of such systems has grown steadily.
However, the goal of simplification has only been achieved in part. This is because a vision system interface is, to date, still a technical interface: geometries and dot clouds, captured images and associated graphics can’t immediately be linked to a specific defect category or a drift within the inspected process. In most cases, identifying the exact nature of the problem requires experienced personnel. However, companies often have trouble finding such personnel, hence the need to optimize their role (and that of the relative inspection systems) within complex plant ecosystems.
Quality trends by defect. The new ‘Classy’ dashboard. The AI revolution is taking us in a specific direction. Not towards further development of the type of control itself, which already offers superb levels of accuracy and specificity, nor towards more on-process intervention, as SACMI already has these aspects in hand. What AI offers, alongside the still-present technical interface, is an innovative dashboard.
Here, the left-hand side of the screen shows the detectable defect types, grouped into ‘families’ (accidental, technological, decoration-OCR, measures, specific). In the middle, an image highlights the area of the object (in this case the preform) where the defect is located. On the right, instead, SACMI has quality statistics. If an issue arises, a specific warning icon comes on, indicating the exact defect type, the affected area and relative statistics-trends.
Where does the revolution lie? In the ability - unthinkable without the aid of AI - to treat the defect not as a ‘mathematical’ association between an acquired image and certain parameters, but to identify exactly what the defect is in terms of the main defect families and sub-families, reaching further levels of specificity with regard to product type, size and the technological process in use.
With Classy AI, a mosquito in the preform is a mosquito, not a black spot. With Classy AI, an insect that accidentally ends up in the body of the preform is not classified as a small black spot but... well, as an insect! That quite an achievement, both technologically - it took many months’ work to develop the algorithms needed to correctly classify the defects - and in terms of the potential usefulness of the system.
With Classy AI, in fact, even the chief mechanic (who has no specific understanding of vision systems) can cast his eye over the dashboard at the end of a shift. And in the even of an alarm, he’ll understand what it is. It’s also possible to filter the problem so that - if necessary - a specialized technician can be contacted. This also ensures the latter can better manage his time and the quantity and quality of tasks he performs on the system.
Feedback on the process has nothing to do with this. Or does it? Using quality control as part of proactive and predictive plant maintenance is one of the greatest challenges facing the world’s industrial vision system providers, SACMI included. In principle, Classy AI has nothing to do with this type of use of Artificial Intelligence.
However, quite the opposite is true if the human factor once again becomes central to the operator-machine-software ecosystem as part of an integrated 4.0 approach to plant management. Having a user-friendly system that allows vision systems to be controlled effectively will, in fact, be highly useful for directing automatic feedback (which can in any case also be activated by inexperienced operators) in a more efficient, finely targeted way.
Just think of a car. Imagine the oil warning light comes on. Before calling the mechanic one will open the hood and check the oil level. Or, if a light stops working, you’ll try and change the bulb before you assume there’s an electrical fault and call the auto electrician. However, here there’s an additional advantage: with Classy AI SACMI already knows what the problem is, who to call and whether to call.
