RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2
Project

Technonikol automates visual quality control of finished products

Customers: Technoflex

Ryazan; Construction and construction materials industry

Product: Artificial intelligence (AI, Artificial intelligence, AI)

Project date: 2020/09  - 2021/02

2021: Pilot project on automation of visual quality control of finished products

On March 10, 2021, TechnoNICOL Corporation announced that on the basis of the Ryazan plant, Technoflex is implementing a pilot project to automate visual quality control of finished products. A system is being introduced on the production line, which, using artificial intelligence, analyzes the technological process around the clock and reveals deviations in the appearance of the produced web or packaging and marking.

On several sections of the production line of the high-resolution chamber, the web is continuously shot. The obtained video information is processed by a computer using artificial neural network technology, which captures all visual deviations from the reference value. The system is able to "see" the smallest changes in the inspected surface with an area of ​ ​ 1 mm2 or foreign inclusions with a diameter of more than 3 mm with a finished product speed of at least 1.6 meters per second.

File:Aquote1.png
We produce materials of the widest nomenclature with great variation of surfaces. For example, we have more than one and a half dozen types of protective films with a logo, which the system must pass further without stopping the line, but in case of a film defect, it must stop the release of material and give a sound signal, "comments Sergey Sukhoruchkin, head of the technical department of the Bitumen Materials and Granules direction of TECHNONIKOL Corporation. And if a person can, without thinking at the subconscious level, determine any changes in the appearance of the finished product, then the system needs to "explain" - describe and classify each defect, translating it into a number. The training of the artificial neural network is continuous, and the larger the sample processed by humans, the faster and more accurate the neural network will analyze the images.
File:Aquote2.png

The system stores in the database a photo, date, time, material name, shift number for each detected deviation and archives it.

If a non-conforming event is detected at the site, the monitoring system interface will notify the personnel using sound and light indicators. If more than 70% of the detected deviation is detected, then a signal is sent to the winding machine relay for instantaneous termination of the web winding in the roll and continuous warning of personnel about the line stopping is activated.

After the winding machine, the product moves to the next section, where the roll passes visual inspection for compliance with the packaging and labeling standard. At this stage, when deviations are detected, in addition to light and sound signals, the control program receives a command to prohibit the product movement along the line until the mismatch is eliminated.

With the successful implementation of the pilot project, TECHNONIKOL intends to begin scaling the technology at the plants of the Bitumen Materials and Granules direction in other regions of the Russian Federation.