Konferencja: ISPEM 2023 – Advances in Production
Aleksandra Rzeszowska, Leszek Jurdziak, Ryszard Błażej, Agata Kirjanów-Błażej
The article presents the application of two unsupervised learning methods, clustering and SOM (self-organizing maps) analysis, for grouping failures in conveyor belts based on data from the DiagBelt + magnetic system. Both methods were used to automatically group the obtained damage signals in order to enable their further identification. Data grouping was performed based on real signals obtained from scanning conveyor belts with the DiagBelt+ diagnostic system. The belt loops on eight conveyors in one of coal mines in Poland, with a total length of 18 472 m, were examined. In total, 96 belt sections from various manufacturers, including belts after refurbishment and at different ages, were studied. The conducted research indicates the effectiveness of using SOM analysis and clustering for grouping damages in conveyor belts based on data from the magnetic system.