Analysis of lobed gas lubricated bearing using neural networks

Abstract

Purpose

The purpose of this paper is to investigate the performance of noncircular five lobe gas lubricated bearings, making use of the efficiency and simplicity of artificial neural networks (ANNs). The effects of different parameters such as compressibility number, mount and tilt angle on static and dynamic characteristics of such bearings are studied.

Design/methodology/approach

For this purpose, various three‐layer neural network models, using Levenberg‐Marquardt method, are selected for training.

Findings

The results obtained as neural network outputs compared with those reported results from finite element method (FEM) for two, three and four lobe journal bearings, are very close. The results for five lobe journal bearing show that the effect of tilt and mount angles on the stability of the bearing system are marginal, while low compressibility number can have more influence on the performance of such bearing systems.

Originality/value

The paper shows that for the performance analysis of gas lubricated journal bearing systems which are cumbersome, due to nonlinearity of their pressure equation, ANNs can be used effectively.

Citation

Rahmatabadi, A.D., Dehghanizade Baghdadabadi, M. and Almodarresi, S.M. (2013), "Analysis of lobed gas lubricated bearing using neural networks", Industrial Lubrication and Tribology, Vol. 65 No. 1, pp. 12-18. https://doi.org/10.1108/00368791311292765