The analysis of the effects of surface texture on the capability of load carriage of journal bearings using neural network

Abstract

Purpose

This paper investigates the load carrying capacity of the journal bearings with steel shafts with varying surface texture in varying revolutions using experimental and neural network (NN) approach.

Design/methodology/approach

In this study, we used a shaft with smooth surface with the same material properties compare their load carrying capacities of the shafts with three different pitches and two different profiles. The experimental data, such as pressure and oil temperature, are employed as training and testing data for NN. Quick Prop algorithm is used to update the weight of the network during the training.

Findings

The designated NN has superior performance for modelling of the system. Therefore, the proposed neural predictor would be used as a predictor for possible experimental applications on modelling bearing system.

Research limitations/implications

Mobil 0W‐40 lubricant was used and kept at temperature of 18°C. The surface of the shafts has been in two types: smooth, that is without and with profiles, that is trapezoidal and saw.

Practical implications

Owing to the parallel structure and fast learning of NN, this kind of algorithm will be utilized to model other types of bearing systems.

Originality/value

Instead of traditional methods, NN has fast learning and parallel processing structure. Moreover, NN can be used to process multiple‐input/multiple‐output data unlike other empirical modelling tools which can map one dependent variable at a time. Therefore, this method is able to predict the load carrying capacity with steel shafts with varying surface texture in varying revolutions satisfactorily where common techniques have failed.

Citation

Si˙nanoğlu, C. (2005), "The analysis of the effects of surface texture on the capability of load carriage of journal bearings using neural network", Industrial Lubrication and Tribology, Vol. 57 No. 1, pp. 28-40. https://doi.org/10.1108/00368790510575969