Specialists from Novosibirsk State Technical University (NSTU) have created an intelligent system for quality control in industry using artificial intelligence, achieving an accuracy of 87%. It is designed for automatic detection of cracks, dents, and corrosion on metal surfaces. The system analyzes photographs taken with a regular camera.
The developed system is based on a triplet neural network that does not need thousands of ready-made images. The system works effectively with a small number of examples for training - a few photos of each type of defect are sufficient for analysis, even if they are taken in poor lighting and at different scales.
According to the developers, the new system easily adapts to rare types of damage without the need for expensive and lengthy data relabeling. It can identify defects, rather than just memorizing images, which makes it unique.
The technology can be used in quality control and predictive maintenance systems in industrial enterprises, especially in metallurgy and mechanical engineering. It is capable of automating the monitoring of steel surfaces, predicting the need for equipment maintenance based on early signs of wear, and increasing the overall reliability and safety of production processes. In the future, the system can be adapted to monitor the condition of bridges, pipelines, and other critical structures.
Read more materials on the topic:
- Neural networks adapted in Moscow to predict defects in steel for nuclear power plants
- AI algorithms will help create unique devices for inspecting large equipment
- Unique technology for printing parts from four metals simultaneously developed at SPbPU
