Abstract
EN 1992-4 permits a fastening to be designed for uncracked concrete only where the absence of cracking can be demonstrated over the whole anchorage depth; in the absence of that demonstration the concrete is taken as cracked. In practice the demonstration is rarely attempted, and the cracked assumption has hardened into a near-universal default. The design consequence is substantial: the characteristic concrete cone resistance in cracked concrete is approximately 70% of the uncracked value, and for bonded fasteners the declared bond resistance in cracked concrete is commonly one half to two thirds of the uncracked value. This paper examines whether the default is proportionate to the evidence supporting it. Five observations are developed. First, both EN 1992-1-1 and EN 1992-4 define the reference crack by its width (0.3 mm) and are silent on its depth, so a shallow surface crack and a crack traversing the full section are treated identically, although their effect on an embedded fastener is not. Second, the guidance issued by the British Board of Agrément (Guidance No. 39) already frames the question probabilistically — as a joint probability involving the location of the crack, the variability of imposed load in time and space, and the independence of the fastening load from that imposed load — a framing that the Eurocode’s binary classification discards. Third, a real flexural crack is wedge-shaped: it is widest at the tensile face and closes to zero at its tip, so a crack of 0.3 mm at the surface is necessarily narrower at every depth below it, whereas the qualification specimen is cracked to approximately uniform width through the member. The two geometries are mutually exclusive. Fourth, the action required to open a crack of 0.3 mm has never been related to the depths at which fasteners are installed; worked examples given here show that the reference crack corresponds to a steel stress of 300 to 350 N/mm², some 70 to 80% of design yield, so that a slab in ordinary service carries surface cracks of 0.14 to 0.22 mm and engages, over a typical anchorage, a mean width under half the value at which the fastener was assessed. For a fastening loaded purely in shear the fastener contributes nothing to opening the crack at all, and none of the three checks that can then govern draws on any parameter obtained from testing in cracked concrete. Fifth, recent experimental work on the contribution of supplementary and surface reinforcement indicates that the cone model underlying the classification omits a resistance mechanism that is systematically present in real members. A worked example from a current European Technical Assessment shows the same pattern in the pry-out check, where the declared product-specific factor places the pry-out resistance permanently above the fastener’s own steel resistance, so that the failure mode the factor quantifies cannot be produced by the product it describes. The paper also notes, as a limitation of the evidence base rather than as an allegation, that the great majority of anchor testing is funded by manufacturers whose commercial interest is served by the more onerous classification. In place of the binary default a graded framework of factors is proposed, ordered so that the condition of the member decides whether a crack of the reference kind can be present at all, the geometry of the anchorage decides how much of it the fastener would engage, and the properties of the connection modify probability and consequence without ever deciding the classification on their own. Its boundaries are stated explicitly: seismic design, and any case of genuine doubt, remain cracked.