ASSESSMENT OF THE LUBRICATION CONDITION OF ROLLING BEARINGS USING CLASSIFICATION ALGORITHMS
Keywords:
Classification, machine learning, Support Vector Machine (SVM) method, Random Forest Classifier (RFC), k-nearest neighbors (KNN), accuracy, bearing, lubricationAbstract
The purpose of this work is to solve the problem of unscheduled failures of rolling bearings installed
on industrial equipment as a result of their improper maintenance during operation. It is known that up to
50% of all unscheduled downtime of industrial equipment occurs due to bearing failure. In this case, the
main reason for bearing failures is violations of the lubrication regime of the rolling elements: excessive
and insufficient quantities of lubricants. These reasons account for up to 36% of the total number of bearing
failures. During equipment operation, it is very difficult to identify and prevent all problems with bearing
lubrication, due to the wide variety of factors influencing their occurrence. Therefore, an urgent task
for research is the development of an automated recommendation system for managing the maintenance of
industrial equipment, with control of the lubrication of bearing units. The paper discusses a method for
classifying the states of bearings depending on their diagnostic parameters: indicators of vibration velocity,
vibration acceleration and temperature. For this purpose, classical machine learning algorithms are
used: KNN, RandomForestClassifier and SVM models. For each model, hyperparameters are determined
to achieve maximum results during training. In the process of conducting the study, an analysis of the
influence of each of the diagnostic parameters - signs on the performance of the classification model was
carried out. Understanding which indicator of bearing performance will be the most important will allow
you to choose equipment condition monitoring devices at a manufacturing enterprise consciously, to solve
specific production problems. The developed algorithm allows us to qualitatively, with 98% accuracy, assess the lubrication condition of rolling bearings and issue recommendations for timely maintenance of
equipment. The classifier model is planned to be used as part of a complex for monitoring the technical
condition of equipment, expanding diagnostic capabilities: in addition to information about the probability
of equipment failure and predicted service life, the diagnostic complex, combined with the proposed
model, will allow influencing the mileage of bearings by improving the quality of their lubrication.
