Automating quality inspection with smart software

The inspection process is however tedious when quality goes down to micron accuracy. It gets increasingly more difficult and it requires an operator to use microscopes for a good look. The question of our customer was: how can we add the intelligence of an operator’s assessment, to an automated inspection process?

Automating 15 experienced operators with software

The current production process involved manual quality inspection by 15 experienced and trained operators. These operators are trained for products that run over a 1,000 different product variants. Automating this inspection process is a great challenge. So there was need for a smart and robust solution.

Two inspection types for optimal quality

When talking about inspection automation, roughly two types of inspection can be distinguished. The first type of inspection is for physical and measurable properties, such as length and diameters. The second type of inspection goes beyond measurable properties, it’s about product quality. Think of deviations on the product’s surface such as scratches, variations in roughness, shininess and other errors. Assessing these properties of the product quality in automation is a real challenge. It’s difficult. Especially when there’s a large product variation in terms of shape, materials and color.

Aiming for full automated inspection brings another challenge to the table. Products that are inspected need to be handled, also after inspection takes place. This means that automated handling can never damage the product, because it simply won’t be detected anymore. Handling the product is thus a very delicate process, which needs to be taken into great consideration when automating production.

To summarize the process steps we integrated

1. Product handling

An industrial robot with special grippers handles delicate products during the production process. We developed a special set of grippers for this solution, making it possible to inspect every product variant. The grippers assure prevention of damages and prints on the product.

2. Image Acquisition

During the production process, three types of images are taken. An image of the upper side, the bottom side and several simultaneous images of the product 360 degrees around. For a reliable result, it is important to generate stable and reproducible images. The hardware (the cameras) and the software (for image processing and deep learning) must communicate in the correct way. Only then, we can assure our customer that a reliable inspection and quality assurance takes place.

3. Image processing

Multiple software packages are used in this step, which processes the images from 3 cameras. We developed a special solution to show the results of all the inspection steps in 1 overview screen. Using a software shell that makes it possible to show the results from the hardware and software systems combined into one intuitive Graphical User Interface (GUI).

4. Deep learning

After inspection, the machine needs to be able to use the results for a subsequent inspection of the same or a different product variant. That is why our system uses deep learning. By using algorithms that search for deviations in the images and comparing these with images of products with normal variations and products with known damages, the system can learn itself to recognize more and more defects. Becoming even more experienced and precise as an operator.

Best and most consistent product quality possible

The automated inspection solution is a stand-alone machine. We integrated this solution into an unmanned production line, that we delivered earlier. This makes production run without interference of operators. The production line handles over 1,500 product variants automatically. Automating the quality inspection of this delicate product resulted in the following benefits:

• Assurance of best possible product quality;
• Consistent high level of inspection;
• Faster quality inspection process;
• Consistent and fully automated repeatable process.

Can your production use a smart manufacturing process?

Are you looking for smart manufacturing processes to improve production drastically? Our experts are here to answer your questions on vision. Send in your questions, as we love to get into your specific challenges.