Sand production in a perforated sample is determined by the onset of a significant discrepancy between strains in two orthogonal directions. The onset is also analyzed by three different sanding models, i.e., shear failure, cohesive tensile failure, and the effective plastic strain (EPS) models, respectively. Comparing these results, we conclude that the results with the shear failure criterion provide the most conservative prediction, and the EPS can provide the closest results to the testing one, given adequate plastic yielding and sanding parameters. Comparing with the cylindrical cases (open hole), both the plastic radius and critical strains calculated for the perforated cavity (cased hole) cases are calculated. A critical equivalent plastic displacement (EPD) is proposed as only one additional parameter is required on top of those from a typical elastoplastic model and such a criterion allows us to determine the sanding onset directly, those using stress, strength or strain whereas can only be indicated or interpreted indirectly, yet difficult to measure in practices.