Probabilistic Analysis for the Performance Characteristics of Engine Bearings due to Variability in Bearing Properties

This paper presents the development of surrogate models (metamodels) for evaluating the bearing performance in an internal combustion engine without performing time consuming analyses. The metamodels are developed based on results from actual simulation solvers computed at a limited number of sample points, which sample the design space. A finite difference bearing solver is employed in this paper for generating information necessary to construct the metamodels. An optimal symmetric Latin hypercube algorithm is utilized for identifying the sampling points based on the number and the range of the variables that are considered to vary in the design space. The development of the metamodels is validated by comparing results from the metamodels with results from the actual bearing performance solver over a large number of evaluation points. Once the metamodels are established they are employed for performing probabilistic analyses. The initial clearance, the length of the bearing, the radius of the journal, and the oil viscosity comprise the random variables in the probabilistic analysis. The maximum oil pressure and the minimum oil film thickness constitute the probabilistic response for which the analyses are performed. The availability of the metamodels allows to compare the performance of several probabilistic methods in terms of accuracy and computational efficiency. A useful insight is gained by the probabilistic analysis on how variability in the bearing characteristics affects its performance.