Benchmark analysis of heat conduction problems with adaptive fea code polyde
Loading...
Date
2017-06
Authors
Jerome Lee Jie Jen
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Adaptive FEM is a topic of interest in research. As different implementations of research code have been developed, it is crucial to test the algorithms on benchmark problems so assessment of the performance can be done in a standard way. This work studies the performance of adaptive FEM code PolyDE on 3-dimensional steady-state heat transfer problem. PolyDE simulations were run for different adaptivity strategies with elements of different polynomial orders on curved surface and flat surface geometry domain. The adaptivity strategies studied were h-FEM and hp-FEM. H-FEM involves refining the mesh at the region where finer elements are required while hp-FEM involves combination of refining the mesh and changing the polynomial orders of the elements to improve the accuracy of the solution. Within the limitation of the number of degree of freedom tested, it was found that error for hp-FEM fits quadratic curves while the error for h-FEM fits linear line in logarithmic-logarithmic plot with respect to number of degree of freedom (NDOF) and CPU time (sec). The goodness of fits was proven with coefficient of determination, 𝑅 2 , which shows value above 0.9 for all the fitted linear lines and quadratic curves. Hence, it was concluded that hp-FEM performs better than h-FEM in faster error convergence. Difference in polynomial order of elements has no significant effect on error convergence rates. PolyDE has issue with memory management. For example, NDOF obtained is limited at 28468 for simulation with hp-FEM starting polynomial order 1 on the curved surface geometry. PolyDE’s performance on curved surface domain geometry is acceptable, with the simulation results for curve surface geometry showing higher error convergence rate than the flat surface geometry. The radial heat sink simulation results replicated in PolyDE is in agreement with [1]. It is concluded that PolyDE is reliable for application in actual LED heat sink analysis. The case studies have been established for benchmarking the performance of PolyDE.