Optimization of Hybrid Journal Bearing using Artificial Neural Network and Genetic Algorithm

Abstract

A rapid and globally convergent predictive tool for optimization of capillary compensated hole-entry hybrid journal bearing having transverse roughness pattern is developed using Artificial Neural Network (ANN) and Multi Objective Genetic Algorithm (MOGA). ANN is trained to compute the objective functions for the multi-objective GA optimization problem of hybrid journal bearing. Five most important parameters such as supply pressure, restrictor design parameter, dynamic viscosity of oil, surface roughness characteristics of opposing surfaces (i.e. variance ratio) and roughness pattern parameter are selected as design variables. Minimum fluid film thickness, frictional torque and critical journal mass of capillary compensated hybrid journal bearing are selected as objective functions. The optimum values of design variables are obtained by interfacing multiobjective GA with ANN. The optimization goal is to maximize the minimum oil film thickness and critical journal mass and to minimize frictional torque on journal surface as three objectives and is achieved using a Pareto optimal front. Achievement of regression value above 99% and close agreement between expected and predicted value of all the three output functions shows the suitability of ANN in Multi-objective Genetic Algorithm optimization problem.