The Application of Nearly Embarrassingly Parallel Computation in the Optimization of Fluid-Film Lubrication©

Abstract

The combination of powerful, yet inexpensive PCs and readily available open sources for parallel computation marks a new era of easy access to massive computation for the tribology community. The study demonstrates the applicability of embarrassingly parallel computation in the optimization of air-lubricated porous bearings with four design variables. To achieve high speedup without increasing the coding complexity, the master computer implements the lattice method to allocate the near-the-same computational load in the master-slave cluster. The effect of master capability on the cluster performance is also presented. The results are compared with that of an unparallelized simplex method and indicate a significant reduction in execution time due to parallelism. In a simulated analysis, a high speedup can also be obtained in dealing with a problem with many design variables. This study provides the framework for optimization of applications with complex tribological models to be solved with minimum execution time.