ABSTRACT
Exceeding the tensile strength is not a sufficient condition to initiate a fracture at a stress concentrator (e.g., a circular hole). Different theories offering a possible explanation are investigated. Neither Weibull statistics, nor a decline in E modulus as a consequence of damage induced near the stress concentrator can offer a satisfying explanation. A fracture criterion taking the stress gradient into account, and introducing a characteristic distance c, allows to cater for a size effect. Applied to a number of Brazilian tests, the characteristic distance seems, especially for small size stress concentrators, to be function of the size of the stress concentrator itself. This is of course inadmissible, and further research will have to reveal if it is not the stress distribution at damage initiation rather than at fracture initiation that should be considered in the yield criterion and secondly if it is not the actual stress distribution rather than an approximation that should be used when averaging the stress along the prospective fracture path.
