. 18.3 (2006): 334-342.
Structural design relies essentially on tests made on cylinders of small size to estimate the probability of failure of prototype members, since full-scale testing of structures to determine their strength is not feasible. The confidence that such scale modeling deserves in terms of representation of actual behavior needs careful examination, due to such factors as material nonlinearities, difficulties of scale representation of particulate materials, and sometimes the impossibility of simultaneously satisfying independent dimensionless parameters. Some failures explained by linear fracture mechanics are associable with strong size effects, as opposed to the cases where small cracks are a material property. Besides research centered on these problems, a number of studies of scale effects have been associated with the increased probability of finding a flaw in larger objects. In fact, geometric similitude may coexist with microscopic randomness of flaws that cause size effects to appear. The type of material of the object under study may also be a decisive factor. For example, scatter of the mechanical properties in unidirectional fiber-reinforced polymers (FRPs) is much larger than in metals due to a larger density of flaws. Thus the strength of FRP laminates may depend on the volume of material involved. Strengthening reinforced concrete columns with FRP wraps leads to new constitutive laws for the overall response of the columns and requires small-scale testing followed by extrapolation for design use. The present paper focuses on the difficulties of this step, based on the experimental data obtained. The questions mentioned above are addressed, and the relevance of the adequate representation of the lateral stiffness of the FRP jacket in the scaled cylinders is emphasized. The paper also addresses the problem of testing confined cylinders with a given slenderness ratio H/D=height/diameter, within the range usually characteristic of short columns, and extrapolating the results for columns of different H/D. The importance of the parameter (thickness of jacket/diameter of column, representative of stiffness of jacket/stiffness of concrete core) is also examined. The influence of the parameter is shown to be relatively minor, whereas the nonscaling of the relative stiffness of the core and jacket would be a major cause of error. The experimental data, in terms of strain and strength, are also compared with numerical models proposed in the literature, and the quality of the approximations is analyzed.
]