【正文】 in accordance with AS . Specimens were cured for 28 days under wet hessian before testing. Table 1 Concrete mix design Material Cement w/c Sand 10 mm washed aggregate 7 mm washed aggregate Salt Slump Quantity 381 kg/m3 9 517 kg/m3 463 kg/m3 463 kg/m3 kg/m3 140 177。 25 mm In order to pare bond strength for the different concrete pressive strengths, Eq. 1 is used to normalize bond strength for noncorroded specimens as has been used by other researcher [8]. (1) 7 where is the bond strength for grade 40 concrete, τ exptl is the experimental bond strength and f c is the experimental pressive strength. The tensile strength of the Φ12 and Φ16 mm steel bars was nominally 500 MPa, which equates to a failure load of and kN, respectively. Experiment methodology Accelerated corrosion has been used by a number of authors to replicate the corrosion of the reinforcing steel happening in the natural environment [2, 3, 5, 6, 10, 18, 20, 24, 27, 28, 30]. These have involved experiments using impressed currents or artificial weathering with wet/dry cycles and elevated temperatures to reduce the time until corrosion, while maintaining deterioration mechanisms representative of natural exposure. Studies using impressed currents have used current densities between 100 μA/cm2 and 500 mA/cm2 [20]. Research has suggested that current densities up to 200 μA/cm2 result in similar stresses during the early stages of corrosion when pared to 100 μA/cm2 [21]. As such an applied current density of 200 μA/cm2 was selected for this study—representative of the lower end of the spectrum of such current densities adopted in previous research. However, caution should be applied when accelerating the corrosion using impressed current as the acceleration process does not exactly replicate the mechanisms involved in actual structures. In accelerated tests the pits are not allowed to progress naturally, and there may be a more uniform corrosion on the surface. Also the rate of corrosion may impact on the corrosion products, such that different oxidation state products may be formed, which could impact on bond. The steel bars served as the anode and four mild steel metal plates were fixed on the surface to serve as cathodes. Sponges (sprayed with salt water) 8 were placed between the metal plates and concrete to provide an adequate contact, Fig. 2. Fig. 2 Accelerated corrosion system When the required crack width was achieved for a particular bar, the impressed current was discontinued for that bar. The specimen was removed for pullout testing when all four locations exhibited the target crack width. Average surface crack widths of , , 1 and mm were adopted as the target crack widths. The surface crack width was measured at 20 mm intervals along the length of the bar, beginning 20 mm from the end of the (plastic tube) bond breaker using an optical microscope. The level of accuracy in the measurements was 177。 mm. Measurements of crack width were taken on the surface normal to the bar direction regardless of the actual crack orientation at that location. Bond strength tests were conducted by means of a hand operated hydraulic jack and a custombuilt test rig as shown in Fig. 3. The loading scheme is illustrated in Fig. 4. A plastic tube of length 80 mm was provided at the end of the concrete section underneath the transverse reaction to ensure that the bond strength was not enhanced by the reactive (pressive) force (acting normal to the bar). The specimen was positioned so that an axial force was applied to the 9 bar being tested. The restraints were sufficiently rigid to ensure minimal rotation or twisting of the specimen during loading. Fig. 3 Pullout test, 16 mm bar unconfined Fig. 4 Schematic of loading. Note: only test bar shown for clarity 10 3 Experimental results and discussion Visual inspection Following the accelerated corrosion phase each specimen was visually inspected for the location of cracks, mean crack width and maximum crack width (Sect. ). While each specimen had a mean target crack width for each bar, variations in this crack width were observed prior to pull out testing. This is due to corrosion and cracking being a dynamic process with cracks propagating at different rates. Thus, while individual bars were disconnected, once the target crack width had been achieved, corrosion and crack propagation continued (to some extent) until all bars had achieved the target crack width and pull out tests conducted. This resulted in a range of data for the maximum and mean crack widths for the pull out tests. The visual inspection of the specimens showed three stages to the cracking process. The initial cracks occurred in a very short period, usually generated within a few days. After that, most cracks grew at a constant rate until they reached 1 mm, 3–4 weeks after first cracking. After cracks had reached 1 mm they then grew very slowly, with some cracks not increasing at all. For the confined and unconfined specimens the surface cracks tended to occur on the side of the specimens (as opposed to the top or bottom) and to follow the line of the bars. In the case of the unconfined specimens in general these were the only crack while it was mon in the cases of confined specimens to observe cracks that were aligned vertically down the side—adjacent to one of the links, Fig. 5. 11 Fig. 5 Typical crack patterns During the pullout testing the most mon failure mode for both confined and unconfined was splitting failure—with the initial (pretest) cracks caused by the corrosion enlarging under load and ultimately leading to the section failing exhibiting spalling of the top corner/edge, Fig. 6