Design predictive tool and optimization of journal bearing using neural network model and multi-objective genetic algorithm


In this paper, rapid and globally convergent predictive tool for dynamically loaded journal bearing design is developed. For accomplishment of such an aim, a neural network model of crankshaft and connecting rod bearings in an internal combustion engine is developed as an alternative for the complicated and time-consuming models. Six most important parameters are selected as inputs of neural network. These parameters are: oil viscosity, engine speed, bearing radial clearance, bearing diameter, slenderness ratio and maximum force applied on bearings. Also, some significant parameters are calculated as neural network outputs. These parameters include: all components of friction loss, all components of oil consumption, minimum oil film thickness, eccentricity, oil temperature rise and displacement relative to shell. In addition, an optimum analysis is performed. To achieve such a target, multi-objective optimization methodology is a good approach inasmuch as several types of objective are minimized or maximized simultaneously. The optimization goal is to minimize friction loss and lubricant flow as the two objectives and develop a Pareto optimal front.