Konica Minolta launches Visual Quality Inspection

Konica Minolta has already been able to carry out numerous use cases with several pilot customers.

• Konica Minolta's customer Phoenix, based in the Czech Republic, and pharma logistics customers in Poland and Spain use plastic boxes to deliver products to pharmacies. These boxes are used repeatedly and can therefore become damaged or contaminated. If a customer returns the products because of such shipping boxes, it can be expensive for the logistics company. VQI inspects the boxes and commands the separator in the production line to remove defective boxes from the line or send dirty boxes for immediate washing.
• Ronal, a car-rim manufacturer based in the Czech Republic, was also looking for a solution for performing visual quality control on rims. Konica Minolta's VQI now detects scratches, dirt, oxides, pores, depressions, burrs and deformations on semi-finished products. It also serves as pre-filter by sorting out products with severe defects that are over one millimetre in size.
• With Severotisk from the Czech Republic, Konica Minolta has been able to win an industrial printing company as a customer: The company has repeatedly had problems with the quality of the paper on its rotary heatset machines, which often led to the machines coming to a standstill. VQI detects various defects in the paper and immediately notifies the staff, preventing the standstill before it happens.
• Another Czech customer, NC Line, which offers complete sheet metal processing services, explored using VQI for detecting defects on metal and plastic parts as part of an innovation project in cooperation with Brno University of Technology.
• Konica Minolta also has a customer in the Netherlands that produces cheese slices for the food industry. Cheese slices produced by cheese dryers can come out of the machine broken or wet, which can affect the quality of the product. The slice can also be misplaced in the packaging. VQI can detect anomalies in the production of cheese slices and immediately informs the operator.
• In Poland, Konica Minolta has an automotive customer that manufactures heavy semi-trailer frames. This production process can be dangerous, especially if the frames have to be moved using a crane. In the plant, several injuries and dangerous situations occurred as the frame was being lifted, usually due to the crane head not being properly attached to the frame — a situation caused by human error. VQI checks that the crane head is properly and safely attached to the trailer frame, thus increasing safety in the workplace.
• In the Czech Republic, Konica Minolta has a customer that manufactures insulation foils. VQI can detect defects in insulation foils better and faster than humans, improving manufacturing efficiency and delivering high-quality products to customers.
• In Denmark, Konica Minolta has a customer that manufactures insulating walls. In practice, the glue often seeps through the final layer: However, this cannot always be seen immediately with the human eye as the glue is transparent. It is essential to detect this defect before the cutting process. VQI detects glue leaks automatically.
• A global appliance producer was looking for a solution for performing visual quality inspections on its washing machine production process in Poland. VQI checks that all of the bolts used in the washing machines are present before the top cover is mounted. In addition, a quality check is carried out with regard to the print quality and readability of the barcode.

Since the algorithms on the FORXAI platform can be specifically adapted to the customer's needs, the aforementioned application examples can also be expanded to include any number of further use cases. As the number of use cases increases, along with the amount of field data and customer feedback, Konica Minolta is refining its algorithms, enabling the AI training programmes to be improved.


[1] Deep Learning learns from its experience and optimises the algorithm each time to improve the result. As with a human brain, neural networks are created, with different layers providing more and more information, resulting in more and more accurate results.