A Parallel Solver for the 3D Simulation of Flows Through Oil Filters

Self Cite

Nowadays, it is widely recognized that computer simulation plays a crucial role in designing oil filters used in the automotive industry. However, even a single direct simulation of the flow usually requires significant computational resources. Thus, it is obvious that solution of optimization problems is only feasible using parallel computers and algorithms.In this paper, we present a general master-slave parallel template, which was specially designed for the easy integration of direct parallel solvers into a parallel optimization tool. We show how an already existing direct solver for the 3D simulation of flow through the oil filter is integrated into our template to obtain a parallel optimization solver. Some capabilities and performance of this solver are demonstrated by solving geometry optimization problem of a filter element.

Section: Direct Problem Solvermentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

See 1 more Smart Citation

A Parallel Solver for the Design of Oil Filters

Starikovičius

Čiegis

Iliev

2011

Self Cite

“…It is well known that for such type of problems the main part of CPU time is spent in solving the systems of linear equations (see, e.g. [6,14]). …”

Section: Parallel Version Of the Discrete Algorithmmentioning

confidence: 99%

Parallel Numerical Algorithms for Optimization of Electrical Cables

Èiegis

Meilūnas

et al. 2008

Self Cite

In this paper we propose new heuristic numerical algorithm for determination of the optimal wires diameters in electrical cables. Two multilevel parallel versions of the optimization algorithm are constructed. The first algorithm is based on master-slave technique and the second algorithm uses the data-parallel strategy. Multilevel structure of the algorithms gives a possibility to adapt them to parallel architecture, for example, cluster of multicore computers. Some results of numerical experiments are presented which agree well with theoretical analysis.

“…Moreover, in order to do tests on an uniform communication network, only one process on a node was used. Up to four processes could run on a node, causing a mismatch with the theoretical analysis due to the wider intra-node (and intra-core) than node-node communication bandwidth, and, due to the concurrency in using global memory on multi-core nodes (the well-known botle-neck of such type architeture) efficiency of the parallel can degrade if 2-4 processes are run on one node [8,33].…”

Section: Performance Testsmentioning

confidence: 99%

A Finite Difference Method for Piecewise Deterministic Processes With Memory. Ii

Annunziato

2009

Abstract. We deal with the numerical scheme for the Liouville Master Equation (LME) of a kind of Piecewise Deterministic Processes (PDP) with memory, analysed in [2]. The LME is a linear system of hyperbolic PDEs, written in non-conservative form, with non-local boundary conditions. The solutions of that equation are time dependent marginal distribution functions whose sum satisfies the total probability conservation law. In [2] the convergence of the numerical scheme, based on the Courant-Isaacson-Rees jointly with a direct quadrature, has been proved under a Courant-Friedrichs-Lewy like (CFL) condition. Here we show that the numerical solution is monotonic under a similar CFL condition. Moreover, we evaluate the conservativity of the total probability for the calculated solution. Finally, an implementation of a parallel algorithm by using the MPI library is described and the results of some performance tests are presented.