Session: 11-45-01: Applications of Computational Heat Transfer

Paper Number: 94919

94919 - An OpenMP Based Parallel Solver for MLPG Analysis of Heat Conduction 

In large-scale simulations involving complex domains, mesh generation is a quite challenging and tedious task both in terms of time and human effort. In recent years, significant research efforts have been invested in developing mesh-free methods to solve engineering and scientific problems without mesh generation. These methods also overcome the shortcomings of the mesh-based methods, such as element distortion in case of problems involving large deformations and crack propagation. Of these methods, the meshless local Petrov-Galerkin (MLPG) method is noticeable. MLPG method is a promising mesh-free method for continuum problems in complex domains, especially for large deformation, moving boundary, and phase change problems.

 

An increased number of published articles in recent years can be considered a measure of the growing research activity in the scope of the MLPG method. Although, the node processing on the MLPG method is often based on processing approximations on the neighboring nodes. This procedure is very time-consuming compared to the finite element method (FEM) procedure. Due to this, very few attempts have been made to apply it to large-scale industrial problems that typically involve millions (or even billions) of nodes. But its parallelization has to be a simple way to speed up, and it justifies the search for other parallel alternatives. Mesh-based methods successfully solve large-scale continuum problems as different parallel techniques are available and well established for these methods. The parallel implementation of the MLPG method requires considerable further research. Modern computing systems, not only high-performance clusters (HPCs) or workstations but even personal computers/laptops, are also equipped with multiple CPUs.  Thus, it is time for an intensive research attempt in this direction, which can provide considerable enhancement to implementing the MLPG method in large-scale industrial simulations.

 

To implement the parallel code on multiple processors, the sequential code has to be customized to execute on the multiple processors to cooperate with different architectures. Generally, there are three main parallelization approaches according to different architectures. The first one is multicore parallelization, for shared-memory multicore computer architectures, with OpenMP application programming interface (API). The second approach is GPU parallelization, based on CUDA API. The latter approach is parallelization on distributed computers, which is implemented with a standard message passing interface (MPI) library.

 

In the present work, we focus our attention on developing an in-house parallel MLPG code for the analysis of the heat conduction problem. The parallel code has been implemented on the shared memory multicore CPU architecture using OpenMP API. Two-dimensional and three-dimensional linear steady-state heat conduction problems are taken as model test problems. Numerical simulations have been performed to find the critical regions of the serial code. An optimum parallel code is developed based on these findings. Based on performance parameters such as speed up and parallel efficiency, the credibility of the parallelization procedure has been established. This work attempts to develop a robust and fast parallel MLPG solver.

Presenting Author: Krishna Singh Indian Institute of Technology Roorkee

Presenting Author Biography: Krishna M. Singh is a Professor in the Department of Mechanical and Industrial Engineering at IIT-Roorkee. He graduated from IIT-BHU, Varanasi, and obtained his M. Tech. and Ph. D. from IIT-Kanpur. He has worked in academia and industry across the globe (India, UK, Japan). He was Senior Visiting Researcher (HIVIPS Fellow) at Hitachi Research Group, Hitachi Ltd., Japan; Honorary Visiting Lecturer and Post-doctoral Research Assistant at Queen Mary, University of London, UK; and JSPS Fellow and Assistant Professor at Shinshu University, Nagano, Japan. His main interest areas are computational mechanics, fluid dynamics, renewable energy, sustainable development, and CAD of thermo-fluid systems. He is a Fellow of the Institution of Engineers, Senior Member of AIAA, and Member of ASME.

Authors:

Abhishek Kumar Singh Indian Institute of Technology Roorkee
Krishna Singh Indian Institute of Technology Roorkee