In a numerical simulation, wrinkles can be detected by a visual inspection of thedeformed mesh, provided that the finite element discretisation is fine enough toallow a proper capture of the wrinkles. This, in general, makes it cumbersome to proceed with the analysis. Rather, it is desirable to proceed with a selectiverefinement to keep the computational cost low (acceptable). In this context [SEL 00,01a, 01b, 02], [BON 92, 94], [ROD 97], [GEL 98], and more generally in finiteelement simulations [SEL 90, 93, 97], adaptive mesh refinement plays an essentialrole. However, this implies that some kind of discretisation error estimators andwrinkling indicators are used to direct the refinement process [SEL 00, 01a, 01b,02], [AMZ 91], [NOR 97, 98], [BRU 97].In this work, the analysis of Hutchinson and Neale [HUT 85], which consists of formulating the problem within the context of plastic bifurcation theory for thin shellelements and its extension by Neale [NEA 89] to account for more generalconstitutive models, is used. Under a number of assumptions, limitations andsimplifications a simple wrinkling criterion with some restrictive applicability, isobtained. Nonetheless, the results are used to locally define a wrinkling risk factor or simply a wrinkling indicator, which, in turn, is used to detect the zones (elements) to be refined in a subsequent adaptive mesh refinement process.Hutchinson analysis is, unfortunately, limited to regions of the sheet that are freeof any contact. When contact is taken into account the problem is further complicated. Furthermore, given that numerical simulations of complex sheet metalforming involve large scale models, it is obvious that global wrinkling indicatorsfound in the literature - mostly based on eigen value analysis of the global tangentstiffness matrix - should not be used because of their high computational cost. Thisis to avoid over-loading the already time consuming deep drawing simulations.Consequently, an indicator based on the local change of curvatures has beendeveloped [SEL 00, 01a, 01b, 02].In wrinkling prediction analyses, the local curvature and the thickness, amongstother parameters, play a major role and should, therefore, be properly approximatedfrom the finite element meshes at all stages of the computation. In this context, theincorporation of discretisation error indicators and adaptive mesh refinement insheet metal forming processes is doubly important in keeping the computation costlow and allowing a comprehensive wrinkling prediction analysis.In recent publications we have presented numerical results on wrinkling prediction with Adaptive Mesh Refinement [SEL 00, 01a, 01b, 02]. However, in anumber of occasions a recurrent question kept rising: How do the numerical resultscompare with those obtained in experimental testing ? To shed some light on suchcomparison, a hemispherical product has been used as a benchmark and a number of product samples stamped with various blank holder forces (BHF) and drawn todifferent depths to capture the onset of wrinkling, its mode and location.