Development of Parallel Computing of Group-Inching Fortification Method and Its Application in Tribological Optimum Design

The goal of this study is to develop a parallel computing algorithm and program for tribological multiobjective optimization using the Group Inching Fortification (GIF) method. The newly developed many-core processor (Intel Xeon Phi) is to be used as the computing device along with the OpenMP directives (the main-stream paradigm) for parallel programming. To maximize the computing efficiency the bottleneck of the parallel optimization and an effective use of many-core system is of vital importance. The published GIF method was developed for sequential (serial) computing which may require modification to obtain a better parallel efficiency for the intended optimization study. In fluid-film lubrication study, numerical computations are usually required to obtain the solutions of various Reynolds equations and related equations. These equations may consider the compressibility of the lubricant, the variation of viscosity (energy equation involved), and the deformation of the bearing surfaces. If an optimization analysis is needed for a complex fluid-film lubrication analysis, the requirement of computational resource can increased exponentially. Recently, Intel developed the many-core Xeon Phi processors for high performance computing (HPC) to compete with the parallel computing conducted by GPUs (Graphics Processing Units). It is expected that the one of the major players in HPC world will be Xeon Phi processors. The main objectives of this study are: (1) Modify the GIF method for parallel execution to cope with the load balancing issues encountered in the multiobjective optimization. Therefore, the air bearing design problem can be effectively solved; (2) In this study, the many-core processor computing is to be implemented and the effective use of the processor cores is to be investigated. The parallel computing efficiency is to be compared with the results obtained by using the traditional multicore processors.

Project ID:PB10607-1029
External Project ID:MOST106-2221-E182-031