Finite element modeling of geomaterials using digital image processing

Abstract

This paper presents a digital image processing based finite element method for the two-dimensional mechanical analysis of geomaterials by actually taking into account their material inhomogeneities and microstructures. The proposed method incorporates the theories and techniques of digital image processing, the principles of geometry vectorization and the techniques of automatic finite element mesh generation into the conventional finite element methods. Digital image techniques are used to acquire the inhomogeneous distributions of geo-materials including soils, rocks, asphalt concrete and cement concrete in digital format. Digital image processing algorithms are developed to identify and classify the main homogeneous material types and their distribution structures that form the inhomogeneity of a geomaterial in the image. The interfaces of the main homogeneous material types are vectorized to form the internal material geometric structure and sub-regions. The vectorized digital images are used as inputs for finite element mesh generations using automatic mesh generation techniques. Lastly, the conventional finite element methods are employed to carry out the computation and analysis of geomechanical problems by taking into account the actual internal inhomogeneity of the geomaterial. Using asphalt concrete as an example, the paper gives a detailed explanation of the proposed digital image processing based finite element method. The paper further applies the new method to the mechanical analysis of the so-called Brazilian indirect tensile test in rock mechanics and pavement engineering. The numerical results show that this new digital image process based finite element method can take into account the material inhomogeneities in the geomechanical analysis which can have significant effects on the tensile stress distribution along the loading axis of the Brazilian indirect tensile tests. © 2003 Elsevier Science Ltd. All rights reserved.