Improvement of the Execution Time of a Cellular Automaton for Subsidence Prediction

Abstract

The movements and deformations of overburden strata and ground surface, due to underground mining, are complex processes affected by: spatial relationship between locations of mining extraction and deformation, mining geometry, timing of the extraction process, geological composition and geotechnical parameters of overburden and dynamics of mining and deformation processes transferred to the surface. The theories developed to explain the mechanism of void transfer from an excavation, through overburden strata, to ground surface essentially assume two extreme models of the overburden medium: continuum elastic or fragmented stochastic. Previous work suggests that the application of Cellular Automata (CA) models leads to results that are coherent with the classical influence function and stochastic models that are well established in the subsidence prediction practice, and it has been shown that the use of CA models leads to development of new subsidence calculation methods with great potential for practical application. The proposed model uses a large number of cells and has proved to be computationally intensive when the resolution of the underlying data structure is increased. In order to facilitate the generation of higher resolution results we concentrate in determining up to which extent the use of supercomputing facilities could improve the execution time of the model, finding an execution time improvement orders of magnitude faster when using parallel processing compared to a single thread of execution. In this paper a general overview of the computational model is presented. Also, efficiency in resource requirements, memory and CPU time, as well as algorithmic flexibilities are examined.