Summary of Experimental Investigation
A brief summary of the experimental study is presented in this section, whereas more details can be found elsewhere (Shaat and Fam 2006). The experimental study included 5 axial compression tests, conducted on a standard 89 x 89 x 3.2 mm HSS section (Fig. 1(a)) with nominal yield strength (Fy) of 380 MPa. The length of the pin-ended members was 2380 mm, which corresponds to a slenderness ratio of 68. Ultra-high modulus unidirectional carbon fibre sheets were bonded to the specimens in the longitudinal direction. A single layer was 0.54 mm thick and had tensile strength and modulus of 510 MPa and 230 GPa, respectively. A layer of glass-FRP (GFRP) sheet was first installed directly on the steel surface before applying the CFRP layers to prevent direct contact between carbon fibres and steel, which could lead to galvanic corrosion. The GFRP lamina was 1.46 mm thick and had tensile strength and modulus of 855 MPa and 20.3 GPa, respectively. The stress-strain curves of steel, CFRP, and GFRP are illustrated in Fig. 1(b).
(b) Stress-strain curves (c)Overall buckling of specimen
The tested specimens included a control (unretrofitted) specimen and three specimens retrofitted with one, three and five layers of CFRP, applied to two opposite sides in the plane of overall buckling. The fifth specimen was retrofitted with three layers, applied to all four sides of the specimen. The specimens were given identification codes. For example, 3L-2S indicates three CFRP layers applied on two opposite sides of the specimen.
The gain in axial strength of the FRP-retrofitted specimens ranged from 13 to 23 percent. The strength gains, however, did not correlate directly to the number of CFRP layers. As indicated earlier, this was attributed to the variability of geometric imperfections among the specimens, which is possibly due to a slight out of straightness of different values among the specimens, or minor misalignment within the test setup, or a combination of both. In all specimens, failure was mainly due to excessive overall bucking of the specimen, as shown in Fig. 1(c), followed by a secondary local buckling in the compression side, at or near mid length of the specimen. The local buckling took the form of inward buckling of the compression face and outward buckling of the two side faces, which was clearly revealed after testing by cutting the specimen, as shown in Fig. 1(d). For the FRP – retrofitted specimens, the secondary local buckling in the compression side was associated with a combined delamination and premature crushing of the FRP sheets. For specimen 3L-4S, retrofitted on four sides, the CFRP on the sides have also fractured due to the local bending associated with the outward buckling, as shown in Fig. 1(e).
Figure 2 shows the load versus axial strain of all specimens, at the two opposite sides, at midlength. The figure shows that both sides are under compression, up to a certain load, where excessive buckling starts. At this load level, the strain readings at the outer surface revert to tension while strains at the inner surface show rapid increase in compression. The strain gauges on the compression side failed as a result of debonding and crushing of CFRP sheets. By carefully examining the strain readings, an average strain value of 0.0013 mm/mm can be defined as the strain at which CFRP failed in compression.
In order to predict the load versus axial and lateral displacement responses of axially loaded slender HSS steel members retrofitted with CFRP sheets, a non-linear model has been developed. The model accounts for both material and geometric (second order effects) non-linearities as well as residual stresses. An incremental approach is used, where the concepts of equilibrium and strain compatibility are satisfied at each loading step. The stress-strain curve of steel is assumed to follow an elastic – perfectly plastic model, as shown in Fig. 1(b). On the other hand, FRP materials are assumed to behave linearly up to failure. The following sections provide detailed description of the model.