Primena mašinskog učenja za identifikaciju stanja kotrljajnih ležajeva kod poljoprivredne mehanizacije na osnovu vibracija

Abstract

Abstract: Mechanization maintenance represents preventive activities through constant monitoring of working parts of agricultural mechanization and timely intervention only if a certain measure deviates from the established values. Rolling bearings are one of the main components of rotating machines, and their properties directly affect the reliability of agricultural mechanization. The shorter service life of rolling bearings leads to lower reliability of agricultural mechanization, which is the consequence of inadequate technical maintenance and use. Bearing's condition monitoring on machines can be achieved by widely used methods that analyze vibration signals. This paper aims to apply a model obtained by machine learning to recognize the condition of rolling bearings with sufficient accuracy using vibration data. Normal states and states with bearing errors were taken into account so that the realized model could be used for early detection of unfavorable mechanic operation and prevention of major damages. Detecting the faulty conditions of rolling bearings based on vibrations at an early stage would contribute to the timely reaction of users, preventing major breakdowns and economic losses. The idea is to transfer the resulting model to devices within the concept of Fog computing and apply it close to the working machine and the operator.