Fractures, in the form of joints and micro-cracks, are commonly found in natural rocks, and their failure mechanism strongly depends on the crack coalescence pattern between pre-existing flaws. Determining the failure behavior of non-persistent joints is an engineering problem that involves several parameters such as mechanical properties of rock, normal stress and the ratio of joint surface area to the total shear surface area. To investigate the impact of such parameters on crack coalescence, the artificial neural network was applied. By this way, a number of networks of threshold logic units, facilitating with adjustable weights, have been tested. For training process, here the computational method adopted was a back-propagation learning algorithm. In the present paper, the input data considered for crack coalescence are geomechanical and geometrical parameters. As an output, the network estimates the crack type coalescence (i.e. mode I, mode II or mixed mode I-II) that are to be used to analyze the stability of geomechanical structures. The paper measures the network performance and then it compares the results with those acquired through an experimental method. The analysis indicates that the influential parameters on the crack coalescence are the Joint Coefficient (JC) which is the ratio of the joint surface to the total shear surface area, normal stress and mechanical properties of bridge material.
It is well known that the strength of rock mass is reduced mainly by the rock joints. However, the failure in the rock mass, some time, is limited to a single discontinuity alone. Generally, several discontinuities exist at different sizes that constitute a combined shear surface hence; the intact rocks located between neighboring discontinuities, called the rock bridges (Fig. 1), are of a great deal for shear resistance of the failure surface (Einstein, 1983).