Implementation of a Non-Linear Regression Model in Rolling Bearing Diagnostics

Abstract

The reliable operation of a mechanical system is dependent on the condition of installed rolling bearings. This paper presents a measuring system and equipment for analyzing the condition of rolling bearings. The developed system is a modular device (VibroLog) for measuring several diagnostic parameters for bearings, including vibration, number of revolutions, temperature and sound pressure. The device is a data collector and analyzer. Software in the Java programming language for the analysis and presentation of data obtained by VibroLog has also been developed. This paper is concerned with the development of a mathematical model for the evaluation and prediction of the qualitative state of rolling bearings in real operating conditions. It also presents results of measurement and mathematical modelling of the results. The model was formed by non-linear regression analysis as one of the most widely used statistical techniques. The analysis of the experimental data showed that the rate of change of a variable is proportional to its actual value. The model was tested on several rolling bearings having different degrees of damage. The developed model makes it possible to evaluate and predict the condition of the bearing by measuring the sound pressure level, which is simpler than vibration measurement in real operating conditions. During testing, the model generated results within the prediction error limits. The developed vibrodiagnostic system and the created model enable condition assessment and prediction for a wide range of rolling bearings.

Keywords

• diagnostics
• non-linear regression analysis
• prediction
• rolling bearings