An extensive experimental work has been conducted on pre-cracked rock-model materials. Prismatic gypsum specimens have been prepared with three pre-existing closed flaws. All the flaws have a constant length of 12.7 mm and are parallel to each other. Different specimens are prepared by changing flaw angle, spacing and continuity. The specimens are loaded in uniaxial compression until failure. In each test the nature of any new crack (tensile or shear), initiation stress and angle, and coalescence stress and pattern are recorded. Three different types of cracks have been observed: wing cracks, coplanar shear, and oblique shear cracks. These are the same types of cracks observed from open flaws. The most significant difference between open and closed flaws is the stress at which initiation occurs. For the same geometry initiation is systematically higher for closed flaws than for open flaws.
There are two competing mechanisms for failure in rocks and other brittle geomaterials: tension and shear. Due to their lower toughness in tension (e.g. Bieniawski 1967) tensile crack initiation may be favored. However the magnitude of tensile stresses at the tips of pre-existing cracks decreases as confinement increases, and at some point crack initiation in shear prevails (Bobet & Einstein 1998a). Wing (tensile) and secondary (shear) cracks have been observed to initiate at the tips of pre-existing fractures in rock and rock-type specimens loaded in compression (Ingraffea 1985, Petit & Barquins 1988, Huang et al. 1990, Reyes & Einstein 1991, Chen et al. 1992, Germanovich et al. 1994, Shen et al. 1995, Wong & Chau 1998, Bobet & Einstein 1998a, b, Bobet 2000, Wong et al. 2001, Sagong & Bobet 2002, Sahouryeh et al. 2002, Li at al. 2005, Wong & Einstein 2006, Ko et al. 2006).