Application of Graphics Processing Unit Computing for Fluid-Film Lubrication Study

The solution of some form of Reynolds equation is usually required in many fluid-film lubrication analyses. And an effective solution method for the Reynolds equation is of vital importance in pactice. These equations include the nonlinear compressible-fluid Reynolds equation and the generalized Reynolds equation that considers the effect of three-dimensional variation of lubricant viscosity. Because these equations have no theoretical solutions, the solutions for these equations rely on some efficient numerical methods, including the use of parallel computing. Thus, an optimization practice, for example, can then be carried out within reasonable time. The computing based on the graphics processing unit (GPU) has been actively studied and applied recently. Today, the GPU computing hardware and the supporting programming software are ready for engineering and scientific applications. The main advantage of the GPU computing is its ability to handle a large number of concurrent calculations (versus only a handful of core processors available in a multi-core computer). The GPU computing can also be conducted simultaneously with the multi-core computing to increase the overall speedup of the system. The GPU computing is an emerging and powerful tool for lubrication study. In this study, the main focus is to apply the high performance GPU (NVIDIA) to solve the Reynolds equation effectively in parallel. The results are to be compared with the results conducted in a multi-core workstation. In this proposed two-year project, the main objectives are: (1) Carry out the parallel successive-over-relaxation computing in the multi-core workstation, of which the results are to be used for efficiency comparison; (2) Conduct the GPU computing using a similar parallel iterative method for the incompressible-fluid Reynolds equation. Some modification of the method may be required; (3) Apply the GPU computing in solving the compressible-fluid Reynolds equation. The equation has to be linearized before being solved, which requires an efficient parallel computing method for the inner and outer iterative loops; (4) Perform air bearing optimization computation by uniting the computing capability of the GPU and CPU concurrently. The capability of each approach can then be comprehended. The items (1) and (2) are expected to be completed in the first year of the project, and the items (3) and (4) are to be completed in the second year.

Project ID:PB10007-0371
External Project ID:NSC100-2221-E182-033