ARTIFICIAL INTELLIGENCE METHODS IN DIAGNOSTICS OF ROLLING PRODUCTION EQUIPMENT. DIAGNOSTICS OF TENSION LEVELER BEARINGS

UDC 681.518.5
DOI: 10.36871/2618-9976.2023.10-2.005

Authors

Alexander L. Shestakov,
Doctor of Technical Sciences, President of South Ural State University
Pavel L. Kachurin,
Deputy Slitting Lines Shop Manager, Magnitogorsk Iron and Steel Works
Vladimir V. Sinitsin,
Ph.D., Deputy Head of Scientific Research Laboratory for SelfValidating Sensors, Systems, and Advanced Instrumentation, South Ural State University
Ivan I. Fedosov,
Junior Researcher of Scientific Research Laboratory for SelfValidating Sensors, Systems, and Advanced Instrumentation, South Ural State University
Alexei V. Erpalov,
Ph.D., Senior Staff Scientist, Center for Vibration Testing and Structural Condition Monitoring, South Ural State University
Olga L. Ibryaeva,
Ph.D., Senior Researcher of Scientific Research Laboratory for SelfValidating Sensors, Systems, and Advanced Instrumentation, South Ural State University

Abstract

This paper discusses the problem of diagnosing the condition of bearings in a tension leveler of a sheet mill. Due to aggressive conditions and location features, traditional vibration diagnostic methods are not applicable. The paper proposes an innovative solution based on a multizone temperature sensor. To build a model of the normal behavior of the system, a multivariate state estimation model and a neural network of the autoencoder structure were used. A comparison of the models’ performance is provided.

Keywords

Rolling bearing diagnostics, Autoencoder, Multivariate state estimation model, Multi-zone temperature sensor