Influence of service loading and the resulting micro-cracks on chloride resistance of concrete

(cid:1) Effect of microcracks due to sustained loading on chloride transport into concrete was obtained. (cid:1) Stress levels above 50% of the ultimate can cause a great increase in chloride ingress. (cid:1) Mineral additives perform well against chloride ingress in concrete subjected to sustained loading. (cid:1) At high stress levels, removal of the load results in a recovery of the chloride resistance. that chloride migration coefﬁcients changed little when the stress level was below 50% of the f u ; however, it is desirable to keep concrete stress less than 25% f u if this is practical. The effect of removing the load on the change of chloride migration coef- ﬁcient was also studied. A recovery of around 50% of the increased chloride migration coefﬁcient was found in the case of concretes subjected to 75% of the f u when the load was removed. (cid:1) 2016 The Authors. Elsevier Ltd. ThisisanopenaccessarticleundertheCCBYlicense(http://


Introduction
Chloride-induced corrosion of steel in reinforced concrete structures is one of the most serious problems affecting their service life and if unattended it can put the infrastructure at risk and endanger people's lives. An understanding of the processes of chloride transport in concrete is very important for engineers attempting to predict the service life of reinforced concrete structures and numerous studies dealing with this topic can be found in the literature [1][2][3][4][5][6][7][8][9]. However, most of them have focused on concretes which are neither cracked nor subjected to any structural loading. The work by Gowripalan et al. [10] showed that chloride diffusivity of concrete in the compression zone was much smaller than that in the tension zone when concrete beams were subjected to flexure. Win et al. [11] suggested that chloride can readily penetrate into the concrete through an open crack. Other studies on the influence of crack width on chloride transport [12][13][14][15] showed that the influence of micro-cracks on chloride ingress can be significant.
Although it has been recognised that structural cracks do influence the chloride transport and chloride induced corrosion of steel in reinforced concrete structures, there is little published work on the influence of micro-cracks due to service loads on these properties. In some studies [12,[16][17][18][19][20][21] chloride transport tests on concrete have been conducted after the load was removed, but the results of concrete under loading and after the load is removed can be different. Based on tests conducted under sustained compressive loading, Antoni et al. [22,23] reported a significant increase of chloride transport into plain concrete when the compressive stress level was above 50% of the ultimate stress -f u . In contrast, based on a numerical simulation a linear decrease of chloride diffusion in concrete under sustained compressive stress from an unloaded state up to 50% f u was reported by Wang et al. [24]. It was found that some conflicting views exist in the literature on critical stress level and chloride transport properties of concretes under different loading conditions. Further, most of the work focused on concrete containing only Portland cement CEMI [24,25], and data on concrete with mineral additives are still limited.
The existing service life prediction models, such as Life-365 [26,27], AGEDDCA model [28] and DuraCrete [29], do not consider any effect of loading or consequential cracking (macrocracking or micro-cracking; micro-cracks are defined below as cracks <0.1 mm wide), hence the predicted service life may not be reliable for most structures in service. The effect of stresses and micro-cracks on chloride transport is not considered in the current Standards. The Eurocode 2 [30] restricts the crack width to 0.3 mm for reinforced concrete in chloride-laden environments but has no restrictions on concretes with small cracks (<0.3 mm) or micro-cracks (<0.1 mm). As a consequence, premature deterioration of structures in service is common. For example, on the basis of a field survey of 57 bridges in Kansas, USA, Lindquist et al. [31] have indicated that the chloride concentration at a depth of 76 mm from the location of a surface crack can exceed the corrosion threshold amount within the first year. Another survey of 219 marine structures along the Norwegian coastline has shown that the signs of corrosion could be seen as early as 5-10 years, which means that the time for corrosion initiation was even earlier [32]. Almost all of these structures required major repair at the age of 20 years, which is significantly earlier than their design life of 100 years.
Therefore, the influence of micro-cracks due to structural loads on the chloride transport in concrete was investigated and thereby recommendations for any change of service life designs of such structures are made. As concrete structures exposed to chloride environments are manufactured with cementitious materials containing mineral additives, the effect of such cementitious materials on micro-cracking under sustained loading and its consequential influence on chloride transport were part of the research reported in this paper.

Materials
The coarse aggregate used was crushed 5-20 mm well graded basalt (size 5-10 mm 33% and size 10-20 mm 67%) and the fine aggregate was natural medium graded sand conforming to BS EN 12620:2002 [33]. Both types of aggregates were oven-dried at 100 ± 5°C for 24 h to remove the initial moisture content and then allowed to cool down to room temperature ($18°C) in air-tight containers before being used. Class 42.5 N Portland cement CEMI conforming to BS EN 197-1:2011 [34] was used. Ground granulated blast-furnace slag (GGBS) complying with BS EN 6699: 1992 [35], pulverised fuel ash (PFA) complying with BS EN 450 [36] and silica fume (SF) conforming to BS EN 13263 [37] were used to replace part of CEMI in the mixes studied.
The proportions for each mix, along with their physical properties, are reported in Table 1. As can be seen from Table 1, some of the mixes were added with a polycarboxylate polymer-based superplasticiser conforming to BS EN 934-2 [38]. The solid content of the superplasticiser was 29%. The water content of superplasticiser was considered to be part of the mixing water and hence this quantity was deducted from the water content reported in Table 1. Water from the mains at a temperature of 20 (±1)°C was used to manufacture all mixes.

Manufacture of test specimens
Four blocks (250 Â 250 Â 80 mm) and six 100 mm cubes were cast for each mix. The required quantities of the mix constituents were batched by mass and then mixed using a pan mixer in accordance with BS 1881-125: 1986 [39]. The slump was measured immediately after batching the concrete. The concrete was placed in the moulds in three layers and each layer was compacted using a vibrating table until air bubbles appearing on the surface stopped. The cast surface of each concrete specimen was then finished with a trowel and covered by a thick black polythene sheet to prevent any loss of water. The samples were demoulded after 24 h, they were then placed in a water bath at 20 (±1)°C for 3 days, after which they were wrapped in wet hessian and thick black polythene sheet. The specimens were then placed in a constant temperature room at 20 (±1)°C.

Compressive strength test
The cubes were used to carry out the compressive strength test at an age of 28 and 56 days for each mix, the results of which have already been reported in Table 1.

Chloride migration test
Cores of size 100 mm diameter and 80 mm thickness were cut from the blocks at an age of 56 days and the outermost, approximately 15 mm thick layer, was cut off from the two end surfaces using a water-cooled diamond saw. This was done to eliminate the effects of the boundary layer on chloride transport and ensure a smooth test surface of the samples. The middle part of 50 (±1) mm was thus used as the test specimen. The thickness of each specimen was measured with a digital calliper to an accuracy of 0.1 mm. The chloride migration test was carried out using these disks. In order to test them under sustained loading, they were compressed by two diametrically opposite loads by using a special setup that was designed and manufactured, as shown in Figs. 1 and 2. This test setup is novel in the sense that it was not used before and was designed to stimulate micro-cracking of reinforced concrete under the sustained loading condition.
Samples were saturated as per the preconditioning procedure outlined in NT BUILD 492 [40]. The cores were placed in a vacuum container after they were surface-dried, by making sure both end surfaces were exposed. The pressure inside was reduced to 30 (±5) mbar and the vacuum was maintained for three hours. Then the container was filled with Ca(OH) 2 solution to immerse all specimens, with the pump still running. The vacuum was maintained for a further hour before allowing air to enter. The cores were kept in the solution for another 18 h, after which they were taken out for placing in the test setup. The cores were then coated on the curved surface with a silicon coating material and clamped in position (Fig. 2). After the silicon had hardened, the specimens were loaded at a constant rate of 1.23 (±0.1) kN/s until failure. The average load of the three failure loads in each mix was taken as the maximum load and the related stress was taken as the ultimate stress -f u for each mix. The typical crack pattern at the failure of the core specimen is shown in Fig. 3.
Three loading conditions (Control -no loading, under loading and after unloading) were utilised. Fig. 4 shows that cores under different loading conditions were treated separately before testing. The load was removed after 1 min of loading for the After Unloading condition and UPV tests were conducted before the load was applied and after the load was removed. The load was kept constant throughout the test period for the under loading condition, UPV tests were conducted before and after the load was applied. No load applying and UPV tests for the Control condition. Different stress levels (equivalent to 25%, 50%, and 75% of f u ) were applied to the cores by fastening the nuts at the top of the brackets using a torque wrench. Strain gauges placed on bolts were used to determine the forces in the bolts, indicated in Fig. 2. An ultrasonic pulse velocity (UPV) test was conducted before and after the load was applied.
Control: Specimens were ready for testing and no UPV test was conducted. Under loading: Different levels of load were applied and maintained for the tests conducted. An UPV test was conducted before and after the load was applied. After unloading: Different levels of load were applied and kept constant for 1 min, and then the load was released. UPV tests were conducted before applying and after releasing the load.
As shown in Fig. 3, cracks along the direction of loading were formed when the cores were under a certain load. The ultrasonic pulse velocity (UPV) in 3 different directions (Fig. 5) was used to determine the damage degree h, which is the change in UPV [41] as per Eq. (1) caused by the loading. As illustrated in Fig. 5, a direct method and two indirect methods were used to measure the UPV for all mixes. Exactly the same locations were utilised before and after each load was applied and vaseline petroleum jelly was used as a couplant between the transducer and the concrete surface. The transducers with a frequency of 54 kHz were pressed by hand at a constant force against the concrete surface to ensure good acoustical contact. Repeated readings at each location were taken and the minimum transit time was reported. After each test, the concrete surface was polished by means of a sand paper so as to avoid any jelly being left on the surface.
where h is the damage degree calculated from the change in UPV before and after loading; V 0 and V are UPV measured at the same location in the concrete specimen before and after loading was applied; L is the measured distance between the two transducers. All cores were tested for chloride transport in accordance with the NT BUILD 492 test procedure [40]. In the condition of 'under loading', rubber rings and spacers were fitted on the cores to prevent any leakage. The rubber ring also served to prevent any direct contact between the solution in each cell and the metal frame in order to avoid the passage of direct current through the metal frames. The end surface nearer to the as-cast surface was the one exposed to the chloride solution (catholyte). The catholyte reservoir was filled with 10% (by mass) NaCl solution and the anolyte reservoir with 0.3 M NaOH solution (in de-ionised water). The electrodes were connected to the power supply and the test voltage was selected according to NT BUILD 492 [40]. At the end of the test the cores were taken out of the frame and split into two halves along the cracks in the same direction of loading in order to make use of the worst case of chloride penetration for calculating the chloride migration coefficient (highest chloride penetration was expected to be along the crack, but splitting perpendicular to the cracks would have produced different chloride migration coefficients). The chloride migration coefficient was determined according to Eq. (2) [40]. For each test 3 replicates were tested and the average value was determined.
where D nssm (Â10 À12 m 2 /s) is a non-steady state chloride migration coefficient obtained from a non-steady state migration test, U (V) is the absolute value of the applied voltage, T (°C) is the average value of the initial and final temperatures in the anolyte solution, L (mm) is the thickness of the specimen, x d (mm) is the average value of the penetration depth and t (hour) is the test duration.

Results and discussion
3.1. Influence of service loading on D nssm Fig. 6 shows a case of typical changes in D nssm due to the applied load. There is almost no change of D nssm when the stress level increased to 25% f u ; above this level and up to 50% f u , D nssm increases a little; with the stress level increasing up to 75% f u , a significant increase of D nssm is obtained. No visible cracks were observed in any of the specimens after testing.
Mehta and Monteiro [42] have suggested that there is no change in behaviour of concrete up to 30% of the ultimate stress, slight damage to the structure between 30% and 50% of the ultimate stress, significantly increased damage between 50% and 75% of the ultimate stress, and the system becomes unstable beyond 75% of the ultimate stress. The changing behaviour of D nssm of 0.65-CEMI concrete presented in Fig. 6 agrees well with the crack developing stages of CEMI concrete under compressive loading suggested by Mehta and Monteiro [42]. Below 25% f u is within the elastic range of concrete, and there is no development of microcracks. Around 50% f u , there is a stable development of microcracks in the interfacial transition zone. Up to 75% f u , cracks begin to form and increase in the matrix and the crack system in the interfacial transition zone becomes unstable. Sustained loading above 75% f u should be avoided because the crack systems of concrete become extremely unstable and can easily fail [42].
As shown in Fig. 6, a significant difference can be found between the values of D nssm in concretes before and under loading. Concrete structures have to sustain service loading, such as live and dead loads; therefore, the stress level is an important factor to consider in the context of chloride ingress. The effect of loading on chloride ingress into concrete can be considered in the design stage of reinforced concrete structures if the range of the actual stresses on the structure is known.
Typical relationships between characteristic values and design values of the concrete strength and the stress are shown in Fig. 7    for accidental situation [43]. Then the maximum design stress level of in-service concrete can be calculated as in Eqs. (4) and (5).
Persistent and transient (normal loading condition): Accidental (unusual loading condition): From the above calculation, the normal stress level is around 57% f u and the accidental stress level is around 71% f u , as shown in Fig. 8. Stress levels discussed here are related to design loads; in service, real loads are lower due to the use of partial safety factors in design. Further, real loads and stresses also depend on live to dead load ratios.
It can be seen from Fig. 6 that D nssm of concrete, within the range of stress levels 57% and 71% f u could increase by 100-200% of that without loading. Thus the service life predicted, based on concrete in unloaded state, might be too optimistic, which suggests that the actual service loading should be considered for the durability design stage.

Influence of mix variables on D nssm
3.2.1. Influence of w/b on D nssm Fig. 9 shows the relationship between D nssm and the stress level for each concrete mix. The D nssm values might be lower than some of the published data, but the overall quality of the data is good and the difference between mixes is clear. As one would expect, the w/b has a significant influence on D nssm for each type of concrete (i.e. binder type). The value of D nssm of 0.45-CEMI and 0.45-10%SF is much lower than that of 0.65-CEMI and 0.65-10%SF respectively. This is because concrete with a higher w/b has a higher porosity as well as higher percolation. Fig. 10 shows the relationship between D nssm and the actual value of the load applied for each concrete mix. The change in chloride migration coefficient for each mix due to the application of 43 kN (highest load that was resisted by the 0.65-CEMI concrete) is reported in Table 2. The load 43 kN was selected because D nssm for all mixes are available under this load. Also reported are the changes in chloride migration coefficient due to approximately half of 43 kN (ffi20 kN) applied to the samples. The highest increase of 208% is for the 0.65-CEMI concrete subjected to 43 kN. In comparison, the increase for the 0.45-CEMI concrete at 43 kN load was only 22%. This is because concrete with a lower w/b has a higher strength (and hence higher tensile resistance) and can withstand higher load before D nssm begins to increase.

Influence of types of binder on D nssm
As shown in Fig. 9, the curves of D nssm under different stress levels for both PFA and GGBS concretes are quite similar, but they are different from that of the CEMI concretes. The addition of PFA and GGBS resulted in an almost no change in D nssm . The SF concretes on the other hand show a trend similar to that of CEMI concretes, but the overall values of D nssm are greatly decreased. These results suggest that the addition of PFA, GGBS and SF is beneficial to control the chloride ingress for concrete structures under sustained loading. From Table 2, it can be seen that at 43 kN sustained load, 0.65-CEMI concrete has a very high increase in D nssm and both 0.65-35% PFA and 0.65-10%SF concretes have lower increases. However, under the same load, 0.65-35%GGBS and 0.65-65%GGBS concretes had a very low increase and even a decrease in D nssm . The data in Table 2 were calculated based on curve fitting done to the measured data because actual data points were not available for both 43 kN and 20 kN for all mixes except 0.65-CEMI. As a result, no attempt is made to test for statistical significance of the effects reported in Table 2. D nssm of 0.65-CEMI increased rapidly with increasing force applied due to relatively lower strength of this concrete. However, there was little change for 0.65-35%GGBS and 0.65-65%GGBS concretes with the same range of load applied. Thus, the binder type is an important factor to consider for the durability of concrete under service loads.

Model of D nssm
In structural design, the effect of uncertainties of loads is taken into account by using a partial safety factor. A similar approach can be used in the case of chloride transport as well. From the above analysis, the D nssm of concrete under loading can be described as D nssm0 (without loading) multiplied by a factor of k L , as shown in Eq. (6). k L is related to w/b, stress level (f/f u ) and mineral additives (ma). The work of Buenfeld and Okundi [44] suggests that the effect of binder content (cc) should be included in such modelling. Thus k L can be described by a function given in Eq. (7).
Note that k L will always be P1. k L < 1 should be taken as 1 for safety reasons. The data reported in this paper is not sufficient to establish the function k L in terms of the parameters in Eq. (7).  Table 3. Changes below 20% are not considered to be statistically significant and this suggests the load levels below 50% f u do not have a significant effect on chloride penetration resistance of concrete. It can be seen clearly from Fig. 11 that there is a significant recovery after the load was removed. The change of D nssm at 25% f u (elastic range of concrete) was fully recovered when the load was released, but only a part of the increase in D nssm due to    sustained load was recovered when the load was P50% f u , which might have been caused by the presence of residual cracks [42]. The increase and recovery of D nssm for each mix after unloading from 50% f u and 75% f u are compared and summarised in Table 4.

Effect of unloading on D nssm of concretes
D nssm significantly decreased after the load 75% f u was removed and the recoverable part is about 50%. This suggests that the short-time loading might not cause a lasting detrimental effect on the chloride penetration resistance of concrete structures. The non-recoverable part might increase with the frequency of loading when concrete suffers cyclic loading [20,45], which is an important topic to consider for structures subjected to traffic, wind and earthquake loads.

Service life of concretes for different stress levels
Life 365 v2.1 [27] was used to calculate the service life of the reinforced concrete element under assumed different stress levels. A square reinforced concrete column with a cover depth of 60 mm was assumed to be exposed in the tidal zone of the North Sea at Dornach in Scotland, where the data of the temperature and surface chloride content was collected [46]. It was assumed that the Table 3 Percentage change in chloride migration coefficient for each mix due to sustained loading and unloading (compared with no load condition).
Mix group 50% f u 75% f u  load applied to the column resulted in the same level of stress in the concrete cover as in the interior. Similarly, it was assumed that the effect of a short term load (as studied) is the same as of a sustained load (as in the field). The time needed to build to maximum surface chloride concentration of 0.95% (weight of concrete) was assumed to be 5 years [46]. The reported diffusivity reduction factor (m) in literatures for concrete with 30% PFA was 0.66-0.7 [3,47] and for concrete with 60-70% GGBS was 0.621-1.23 [48]. The max-     imum value of m in Life 365 is 0.6 [27], thus, values of m for different types were assumed as: m = 0.5 for 35% PFA concretes, m = 0.6 for 65% GGBS concretes, m = 0.4 for 35% GGBS concretes and m = 0.2 for CEMI concretes. The default hydration time (25 years) and critical chloride content (0.05% of concrete by weight) were used. Only the corrosion initiation time was considered for the service life calculation. There are limitations on these assumptions. One limitation is to relate the result of discs to the columns on site because of their different stress distributions. Fig. 12 shows significant differences in service life predicted between concretes without loading and under loading. Service life of concretes under loading at above 75% f u could be less than half of that without loading. In all the 8 mixes, 0.65-65%GGBS concrete has the best performance with the longest service life at 25% f u . By adding GGBS, the service life of 0.65-65%GGBS at 75% f u is still much longer than that of any other mix under loading or without loading, which suggests that adding mineral additives might be a better way to control the detrimental effect caused by loading level in durability design of concrete structures.
A simplified model of the service life of reinforced concrete under different loading conditions based on the data of this research is shown in Fig. 13. The chloride diffusion coefficient changes with time. It may decrease in the first one or two years due to continuous hydration [3,47,49], but then may increase with time because of micro-cracks caused by the service loading. The assumed service lives of reinforced concretes (0.65-CEMI) in different loading conditions are indicated in Fig. 13. As can be seen, it is not appropriate to predict the service life of reinforced concrete without considering the effect of loading, which is not the current practice. High levels of loading (>50% f u ) may significantly reduce the service life of concrete structures.
3.6. Relationships between D nssm of concrete with no loading, under loading and after unloading Fig. 14 shows the relationship between D nssm of concrete under loading and no loading. All types of concrete show similar behaviour with a higher variation at 75% f u . D nssm of concrete under 25% f u is almost the same as D nssm of concrete with no loading, D nssm of concrete under 50% f u is around 1.43 times of that with no loading and D nssm of concrete under 75% f u is around 2.24 times of that with no loading. D nssm of concrete under loading can be estimated from these relationships and adjustments to the service life based on no-load condition can be made.

Relationship between normalised D nssm and damage degree measured by UPV method
Figs. 15-17 show the relationship between damage degrees measured in 3 different directions (Fig. 5) and normalised D nssm . D nssm /D nssm0 , where D nssm0 is D nssm of concrete without loading, was taken as normalised D nssm . In Fig. 15, the damage degrees obtained from the UPV measurement along the cracks (1-1 in Fig. 5) are plotted against the normalised D nssm . As is evident from this figure, there was no relationship between them for the range of concretes tested for the two loading conditions. A similar finding was obtained in Fig. 16, which shows the damage degrees based on UPV measured parallel to the cracks on the test surface (2-2 in Fig. 5) and the normalised D nssm . However, in Fig. 17, there is an exponential relationship between the damage degrees based on UPV measurement perpendicular to the crack on test surface (3-3 in Fig. 5) and normalised D nssm . These differences can be interpreted by the directions of measurement compared with the direction of cracks. The damage degree can be determined from UPV measurements only if the direction of measurement is perpendicular to the direction of cracks.
The negative values were mainly caused by testing errors, which include the accuracy of test locations and the contact between sensors and test surfaces. However, the negative values were not found at higher load levels, which suggests the sensitivity of UPV test increases with increasing width of cracks. The results of the three figures suggest that the test direction of UPV method should be carefully chosen in order to detect the cracks.

Conclusions
From the above results and discussions, the following conclusions have been drawn: (1) Compressive stresses in concrete can increase the ingress of chloride. Design stresses below 50% f u do not detrimentally reduce the concrete's resistance to chloride ingress. However, overloading should be avoided because it can cause significant increases in the ingress of chloride ions into the concrete. Concrete in aggressive environments with service loading of over 50% f u can have a 2-4 times higher chloride migration coefficients compared to concrete without loading. (2) Significant difference can be found in chloride resistance of concrete under loading and after unloading (load was removed). The chloride migration coefficient of concrete under loading at 75% f u can be reduced by half if the load is released. This indicates that there are two different situations of chloride ingress in concrete which are under live load as well as under cyclic load. Accidental loading is therefore unlikely to cause a lasting detrimental effect. (3) Micro-cracks caused by service loads should be taken into account during durability design and service life prediction. The chloride migration coefficient of concrete under loading is different and can be significantly higher than that without loading. Thus a multiplier should be applied to take account of the effect of loading. The findings of this research have indicated that the afore mentioned factor is a function of water/binder (w/b), stress level, mineral additives (ma) and binder content (cc). Further work needs to be carried out to quantify the relative effect of each of these factors. (4) In the context of all the data presented in this paper, the use of mineral additives such as GGBS would appear to be very effective in extending the service life of structures.