Total Variation Diminishing (TVD) method for Elastohydrodynamic Lubrication (EHL) problem on Parallel Computers

TL;DR

This paper introduces a novel TVD numerical method for elastohydrodynamic lubrication problems, utilizing a parallel solver called PAQIF to efficiently solve large-scale variational inequalities on parallel computers.

Contribution

It presents a new TVD scheme combined with PAQIF for parallel computation of EHL problems, extending to 2D cases with efficient domain splitting.

Findings

Validated the method's accuracy on serial and parallel computers.

Demonstrated improved computational efficiency for large-scale EHL problems.

Extended the approach to two-dimensional cases with effective domain splitting.

Abstract

In this article, we offer a novel numerical approach for the solution of elastohydrodynamic lubrication line and point contact problems using a class of total variation diminishing (TVD) schemes on parallel computers. A direct parallel approach is presented by introducing a novel solver named as projected alternate quadrant interlocking factorization (PAQIF) by solving discrete variational inequality. For one-dimensional EHL case, we use weighted change in Newton-Raphson approximation to compute the Jacobian matrix in the form of a banded matrix by dividing two subregions on the whole computation domain. Such subregion matrices are assembled by measuring the ratio of diffusive coefficient and discrete grid length on the domain of the interest. The banded matrix is then processed to parallel computers for solving discrete linearized complementarity system using PAQIF algorithm. The idea is easily extended in two-dimensional EHL case by taking appropriate splitting in x and y alternating directions respectively. Numerical experiments are performed and analyzed to validate the performance of computed solution on serial and parallel computers.