Numerical studies of paralleling test particles method using statistic independent tests
Journal Title: Технічна механіка, Техническая механика, Technical Mechanics - Year 2015, Vol 2, Issue 2
Abstract
The test particles method (TPM) as a version of the Monte-Carlo statistic method for the stationary state-ment is considered. Nonparalleling a sequential algorithm of the TPM as large independent subproblems (LISs) has been performed at previous stages. Testing the LISs algorithm using the problem of an internal flow through Laval nozzle followed by the environmental spraying jet has demonstrated a high degree of the efficiency and the algorithm acceleration. However, testing with a problem of an internal flow past has revealed LISs significant disadvantages. It has been found that the algorithm developed has violated in fact some TPM principles resulting in poorly determined velocity fields in the course of iterations. A demand arose for construction of a parallel algo-rithm to accelerate the determination of velocity fields. Thus, the algorithm for paralleling on statistically inde-pendent tests (PSIT) has been developed to take into account velocity disturbances in simulating trajectories of test particles on each operational cores in a single formable field. PSIT reduces to a series of parallel tests con-ducted singly on each core. Following computations, the data are synchronized between cores, namely, an ex-change of accumulated variations in velocity disturbances of field molecules. Subsequent drawing trajectories on all of the cores is conducted on a single renewed velocity field. After all of the tests the summary characteristics due to computational meshes are transferred to one of cores where moments of the disturbance function are averaged on the summing time of residence of particles into meshes. Using the PSIT algorithm, test computations are made using a multi-core processor. The results obtained are compared with the analogue data of the TPM sequential algorithm, as well as the one-core computational results. It is established that the number of cores used is not affected the results obtained and affected only the estimated time when using the PSIT. The time for various versions of the PSIT updated algorithm to minimize the estimated time due to reduction of the number of exchanges between cores is fully considered in the paper.
Authors and Affiliations
L. Pecheritsa
Keywords
Related Articles
- EP ID EP355787
- DOI -
- Views 110
- Downloads 0
Aerodynamic optimization of the shape of supersonic compressor stage guide blades
This paper is concerned with the development of approaches to aerodynamic improvement of turbomachine blade rows. The aim of this paper is to verify the workability of the authors’ technique for the aerodynamic optimizat...
Integral-differential equation of fine grinding kinetics
Kinetics for fine grinding the ore is analyzed. The form of an integral-differential Zagustin’s equation of the grinding kinetics is employed to describe variations in the distribution function of particles by sizes in g...
Model for calculating the expected cost of new rocket hardware development with account for uncertainty factors
Rocket hardware represents technically complex expensive and unique engineering systems. At present, de-cisions on the advisability of implementation of new pocket hardware development projects are made based on financia...
Conditions of stability of switching systems
Dynamic systems described by a combination of differential equations and a discrete switching signal are examined. The research objective is to find an upper estimate for the Lyapunov maximum exponent of the switched sy...
Regularities of deformation of elastic massifs
The analysis of the existing solutions for the elasticity theory is demonstration of their serious disadvantages resulted from the application of imperfect physical dependences. In order to eliminate this problem and to...