【正文】 mm bars Fig. 12 Maximum crack width versus bond stress for 16 mm bars Fig. 13 Maximum crack width versus bond stress for 12 mm bars 15 The data show an initial increase in bond strength for the 12 mm specimens with stirrups, followed by a significant decrease in bond, which is in agreement with other authors [12, 15]. For the 16 mm specimens an increase on the control bond stress was observed for specimens with and mm mean crack widths, however, a decrease in bond stress was observed for at the mean crack width of mm. The 12 mm bars with stirrups displayed an increase in bond stress of approximately 25% from the control values to the maximum bond stress. An increase of approximately 14% was observed for the 16 mm specimens. Other researchers [17, 24, 25] have reported enhancements of bond stress of between 10 and 60% due to confinement, slightly higher to that observed in these experiment. However the loading techniques and cover depths have not all been the same. Variations in experimental techniques include a shorter embedded length and a lower cover. The variation on the proposed empirical relationship between bond strength, degree of corrosion, bar size, cover, link details and tensile strength predicted by Rodriguez [24] has been discussed in detail in Tang et al. [28]. The analysis demonstrates that there would be an expected enhancement of bond strength due to confinement of approximately 25%—corresponding to a change of bond strength of approximately MPa for the 16 mm bars (assessed at a 2% section loss). For the 12 mm bars the corresponding effect of confinement is found to be approximately 35% corresponding to a MPa difference in bond stress. The experimental results (14 and 25%, above) are 60–70% of these values. Both sets of data indicate a relationship showing decreasing bond strength with (visible surface) crack width. A regression analysis of the bond strength 16 data reveals a better linear relationship with the maximum crack width as opposed to the mean crack width (excluding the uncracked confined specimens), Table 2. Table 2 Best fit parameters, crack width versus bond strength Unconfined 12 mm Confined 12 mm Unconfined 16 mm Confined 16 mm Mean crack width R 2 Slope (m) ? ? ? ? Intercept (b) Maximum crack width R 2 Slope (m) ? ? ? ? Intercept (b) There was also a significantly better fit for the unconfined specimens than the confined specimens. This is consistent with the observation that in the unconfined specimens the bond strength will be related to the bond between the bars and the concrete, which will be affected by the level of corrosion present, which itself will influence the crack width. In confined specimens the confining steel will impact upon both the bond and the cracking. Corrosion degree and bond stress 17 It is apparent that (Fig. 14) for corrosion degrees less than 5% the bond stress correlated well. However, as the degree of corrosion increased there was no observable correlation at all. This contrasts with the relationship between the observed crack width and bond stress, which gives a reasonable correlation, even as crack widths increase to 2 and mm. A possible explanation for this variation is that in the initial stages of corrosion virtually all the dissolved iron ions react to form expansive corrosion products. This reaction impacts on both the bond stress and the formation of cracks. However, once cracks have been formed it is possible for the iron ions to be transported along the crack and out of the concrete. As the bond has already been effectively lost at the crack any iron ions dissolving at the crack and being directly transported out of the concrete will cause an increase in the degree of corrosion, but not affect the surface crack width. The location, orientation and chemistry within the crack will control the relationship between bond stress and degree of corrosion, which will vary from specimen to specimen. Hence the large variations in corrosion degree and bond stress for high levels of corrosion. Fig. 14 Bond stress versus corrosion degree, 12 mm bars, unconfined specimen 18 Significantly larger crack widths were observed for the unconfined specimens, pared to the confined specimens with similar levels of corrosion and mass lost. The largest observed crack for unconfined specimens was mm pared to mm for the confined specimens. This is as expected and is a direct result of the confinement which limits the degree of cracking. Effect of confinement The unconfined specimens for both 16 and 12 mm bars did not display the initial increase in bond strength observed for the confined bars. Indeed the unconfined specimens with cracks all displayed a reduced bond stress pared to the control specimens. This is in agreement with other authors [16, 24] findings for cracked specimens. In cracked corroded specimens Fang observed a substantial reduction in bond strength for deformed bars without stirrups, while Rodriguez observed bond strengths of highly corroded cracked specimens without stirrups were close to zero, while highly corroded cracked specimens with stirrups retained bond strengths of between 3 and 4 MPa. In uncorroded specimens Chana noted an increase in bond strength due to stirrups of between 10 and 20% [14]. However Rodriguez and Fang observed no variation due to the presence of confinement in uncorroded bars. The data is perhaps unexpected as it could be anticipated that the corrosion products would lead to an increase in bond due to the increase in internal pressures, caused by the corrosion products increasing the confinement and mechanical interlocking around the bar, coupled with increased roughness of the bar resulting in a