The system guides the user through the training process in a simple way.

© Inspecto

The entire set-up process can be carried out by any employee with machine vision or AI skills in just 30 to 60 minutes without external support. This gives plant managers complete control over their quality assurance, allowing them to immediately investigate any issues with visual inspection of their production equipment and avoid high installation and maintenance costs.

The user switches on the controller during setup and ensures that the field of view covers the area to be inspected. He then inserts a fault-free test sample into the field of view and uses a mouse to mark the 'area of interest', i.e. the area in which the system is to detect faults.

Inspekto S70 reverses the parameters of traditional quality assurance, so to speak: Instead of learning what a defective part looks like, it records what a flawless part should look like in a similar way to a human being. If the inspection system detects a part that deviates from this, the user is informed. This greatly simplifies the detection process, especially in cases where manufacturers do not have a sample product for all types of deviations. The system only requires suitable sample products, which are usually available in large quantities in a production plant.

In contrast to traditional QA solutions, where hundreds or even thousands of defect-free and even defective products have to be recorded, the S70 only requires an average of 20 defect-free sample products. Depending on the complexity of the product, additional sample products may be necessary - the system determines the exact number of sample parts required independently. Furthermore, the user does not need any specialist knowledge of data annotation or labeling.

AI for video sensor optimization

With the help of artificial intelligence, the inspection system then automatically optimizes the evaluation of the video sensor. The integrated AI engine automatically adjusts the lighting and camera parameters to the product being inspected and the ambient conditions in order to obtain the best possible image with perfect focus, depth of field, exposure and dynamic range. These algorithms ensure that the optimum parameters are used in the production plant over the entire service life.

In addition, the system's detection and alignment AI automatically locates the products to be inspected and recognizes their three-dimensional position in space. In this way, the user can easily introduce additional good parts into the process if required. The system automatically recognizes these new defect-free samples even at high workloads and even if they are introduced into the process at different locations or in a different orientation than the previous good parts.

Automatic notification

Once enough sample data has been captured, the system notifies the user and the inspection can begin. The system compares the images captured with the appropriate sample references during setup and looks for discrepancies. The user does not need to set up machine learning parameters such as data augmentation, network topologies or final classification thresholds. These tasks are performed by the system's AI engines.

If the system recognizes features during the inspection that were not covered by the sample parts, it informs the user. The S70 reports products identified as defective to the connected PLC so that an industrial robot, a person or an appropriately programmed pick & place system can remove the faulty part from the production line. If the user does not subsequently classify the detected deviation as defective, the product is sent to the next production steps. In future, the system will no longer mark other parts with the same difference to the sample parts.

Inspekto S70 can also be optimized for sorting: With appropriate programming, defective products can be removed from the production line and sent back to a previous production phase for reworking. Only fault-free products are sent to the next machining processes. This allows manufacturers to maximize the productivity of their systems and minimize waste, as no resources are wasted.

Control at every stage of production

Autonomous machine vision gives manufacturers a high degree of freedom for ongoing operations. Traditional QA solutions are usually designed to inspect only one product at a specific point in the production line. The development, creation, installation and testing can take several months to a year, resulting in long downtimes. Conventional systems typically cost between 50,000 and 150,000 euros and are therefore not suitable for small to medium-sized production sites.

Systems for autonomous machine vision, on the other hand, also offer advantages in terms of cost, which with Inspekto S70 is less than 10,000 euros. It is possible for plant managers to position corresponding QA stations at every connection point of the production line and not just at the end, which means that comprehensive quality control can be carried out after each production step.
In addition, the system can be easily moved from one connection point to the next if required. As the entire installation process takes less than an hour and can be carried out while production is running, there is no downtime.

AMV systems are not product-specific and can therefore be used at a production site or within a production line to check different products.

Personalization via apps

To enable a higher level of personalization for users, Inspekto recently introduced a range of apps that can be implemented on Inspekto S70. One of these apps is 'Inspekto Tracks'. It archives all product images and data captured during the inspection. With the help of an active archive, manufacturers can prove at any time that their products have left production without defects. If defective products are nevertheless put into circulation, a company can trace exactly where in the manufacturing process these defects occurred. This makes it possible to identify and optimize those areas of production where defects occur more frequently.

The use of 'Inspekto Types' allows the system to inspect any number of products at the same location on the production line - dozens or hundreds of different products if required.

And last but not least, 'Inspekto Sequence' gives users the option to preset a specific sequence, for example to inspect an item at different stages of production on a production line. Both apps allow the inspected product profile to be changed in real time and support the rapid changes required in modern production lines.

Authors:
Harel Boren is co-founder and CTO of Inspekto in Tel Aviv, Israel.
Yonatan Hyatt is co-founder and CEO of Inspekto in Tel Aviv, Israel.