**Mechanical and Durability Properties of Concrete with Coarse Recycled Aggregate Produced with Electric Arc Furnace Slag Concrete**

#### **Pablo Tamayo 1, Joao Pacheco 2, Carlos Thomas 1,\*, Jorge de Brito <sup>2</sup> and Jokin Rico <sup>3</sup>**


Received: 19 November 2019; Accepted: 23 December 2019; Published: 27 December 2019

**Abstract:** The search for more sustainable construction materials, capable of complying with quality standards and current innovation policies, aimed at saving natural resources and reducing global pollution, is one of the greatest present societal challenges. In this study, an innovative recycled aggregate concrete (RAC) is designed and produced based on the use of a coarse recycled aggregate (CRA) crushing concrete with electric arc furnace slags as aggregate. These slags are a by-product of the steelmaking industry and their use, which avoids the use of natural aggregates, is a new trend in concrete and pavement technology. This paper has investigated the effects of incorporating this type of CRA in concrete at several replacement levels (0%, 20%, 50% and 100% by volume), by means of the physical, mechanical and durability characterization of the mixes. The analysis of the results has allowed the benefits and disadvantages of these new CRAs to be established, by comparing them with those of a natural aggregate concrete (NAC) mix (with 0% CRA incorporation) and with the data available in the literature for concrete made with more common CRA based on construction and demolition waste (CDW). Compared to NAC, similar compressive strength and tensile strength values for all replacement ratios have been obtained. The modulus of elasticity, the resistance to chloride penetration and the resistance to carbonation are less affected by these CRA than when CRA from CDW waste is used. Slight increases in bulk density over 7% were observed for total replacement. Overall, functionally good mechanical and durability properties have been obtained.

**Keywords:** recycled aggregate concrete; electric arc furnace slags; mechanical properties; durability

#### **1. Introduction**

Approximately 90% of construction and demolition wastes (CDW) are currently going to landfills even though they are potentially recyclable [1]. The use of this waste should be a priority to achieve the sustainable development objectives set by the European Commission, although this action is hindered due to lack of facilities and standards, lack of support from governments or lack of users' confidence [1,2]. The use of CDW as aggregates in concrete production, mostly coarse recycled aggregates (CRA), not only means a saving of natural resources derived from the extraction of aggregate, but also economic savings. Analogously, concrete with electric arc furnace slags (EAFS) as aggregate is based on the use of waste (from the steel industry) that would otherwise be deposited in landfills. In this case, the reduction of CO2 emissions in the processes without taking into account the transport and manufacturing of the materials can be as high as 35% [3]. On the other hand, CO2 emissions induced by concrete crushing are not very different from those generated in the production of natural aggregates [4].

To improve the understanding of the paper, the acronyms used and their meaning are shown in Table 1.


**Table 1.** Acronyms used in the article.

Most CRA are produced by crushing concrete that has ended its service life, i.e., they are composed mainly of natural stone and attached mortar. Typically, CRA are materials with lower density and porosity than natural aggregates (NA) because the attached mortar is less dense and more porous than the natural aggregate that it covers. The average density can be 8% lower and the average water absorption 5–6 times those of the natural aggregates [5,6]. According to the current Spanish concrete standards, the aggregates' water absorption must be less than 5% to be used in structural concrete [7,8]. According to Etxeberría et al. [9], the shape index of NA is 25% and 28% for CRA produced in quarries, although its value depends on the crushing process. Typically, laboratory-produced CRA are made with a single crushing stage (usually a jaw crusher), whilst NA are produced with multiple crushing (primary, secondary and sometimes tertiary). De Brito et al. [10] found that when CRA go through the same crushing process as NA their shape index is expected to be lower than that of NA. The water absorption and the shape index are vital for the calculation of the compensation water to determine the total water/cement (w/c) ratio [11].

There are many studies on the use of recycled aggregate (RA) in the production of structural concrete. Most of them only consider replacement of the coarse fraction of the aggregates, because the fine fraction has a great cohesion and water absorption that make it difficult to control the quality of the aggregates [12] and reduce the workability in the fresh state [13]. The risk of contamination of the finer fraction is also higher [14].

The use of CRA is more common also because it has less porosity and adhered paste. Typically, CRA concrete is around 4%–8% less dense [15–17], although this effect can be the opposite for high density CRA (e.g., CRA based on EAFS). The water absorption of these CRA concretes can be 500% higher than that of NA concrete (NAC) [18,19], although it tends to decrease due to the crystallization of hydration products, depending on the crushing age of the source concrete and curing conditions [20–22]. The fresh workability with CRA is lower than that of NAC for equal w/c effective ratios, so it is important to correct the water content to achieve similar slump without the help of admixtures [16,23–28].

One of the main characteristics of this concrete is the presence of three interfacial transition zone (ITZ). One is between the original aggregate of CRA and the cement paste in the source concrete and it is formed by dense hydrates, and in the case of EAFS concrete is of higher quality than with NA [29]. Another is between the old cement paste and the new paste, and the third ITZ is between the NA of the recycled aggregate and the new cement paste. The two ITZs between recycled aggregate (stone and mortar) and the new paste are where the chemical reactions between both generate loose and pore interfaces [30,31]. These ITZs are thus weaker and limit the mechanical properties of CRA [30]. Concrete with 100% CRA shows a loss in compressive strength with respect to coarse NA at 28 days, for the same effective w/c ratio and amount of cement, from 10% to 37% [9,17,27,32–34], approximately proportional to the replacement level [35,36] and depending on the relative strength of the new paste and CRA [37]. This decrease in compressive strength may make it necessary to use about 5% more cement to achieve the same strength as with NA, thus compromising the cost-effectiveness

and sustainability of CRA [9,12]. From 28 to 91 days, the relative increase in compressive strength with recycled aggregate concrete (RAC) is sometimes greater than with NA due to the hydration of unhydrated cement grains [37]. The splitting tensile strength is typically lower but often not by as much as compressive strength, but there are studies where the tensile strength can even be higher with CRA than with NA [9,38], while the modulus of elasticity can decrease by 15% for 30% replacement and 45% for 100% replacement [35,39]. This reduction is because CRA has a lower modulus of elasticity than NA and due to the increase in the effective w/c ratio to maintain workability constant. Reductions in the modulus of elasticity result in increases in the peak strain of concrete under monotonic compression—[32] reports an increase of 20% for aggregate replacement of 100%. It is also known that the use of CRA but also the new cement paste affect the fatigue behaviour of concrete, reducing the fatigue limit and fatigue life [40,41]. Other authors have shown that the multi-recycling of the concrete that contain CRA is limited and that after three recycling cycles, CRA are mostly composed of mortar and new NA are needed in the mix design [42].

The drying shrinkage of concrete with CRA can be 50% higher [43] than with coarse NA, while chloride ions' penetration can reach up to 150% increases [24,44] due to the high permeability of this concrete. Resistance to carbonation is linked to the porosity of concrete [15] and, therefore, to the porosity of CRA, although it is also strongly linked to the chemical composition of concrete [24]. The high porosity of CRA makes carbonation depths increase between 22%–187% for 100% replacement [24,33,45] in comparison to NAC. Some authors propose the use of more crushing stages to eliminate the attached mortar and thus obtain rounder and less porous aggregates [45,46], the use of acids or heat to disaggregate the mortar of RCA [47,48], thermo-mechanical processes [49,50] or the use of several mechanical systems for on-site processing [51]. Another solution presented by the literature is the use of crushed bricks or steel slag [13,52] to compensate the loss of strength and durability, due to their pozzolanicity. However, these beneficiation techniques add new steps to the aggregate production process and increase production costs and the environmental impact of CRA production.

Currently, there is an increasing trend towards the use of steel slag in concrete [23]. The use of EAFS as CRA in concrete is a novelty (to the best of the authors' knowledge, it has never been done before) and its use is justified by the potential benefits of this aggregate and by the boom of its use in recent years mostly in road pavements or hydraulic structures [53]. The characteristics of the source concrete determine the behaviour of the recycled aggregates concrete. Concrete with EAFS offers an improvement in compressive strength by 50% compared to concrete with NA, a slightly higher modulus [54] and a generalized improvement in durability (low water absorption and permeability) [55,56], whereas the roughness of the aggregates allows improving the quality of the new ITZ of CRA. The quality of the CRA from EAFS concrete will compensate the loss of mechanical and durability properties that more common CRA provides, saving natural resources. This concrete has potential applications in foundations, plain concrete walls, or structures where a high self-weight is important (e.g., radiation-proof structures). The results of the physical-mechanical and durability tests, for concrete with coarse replacements of 0%, 20%, 50% and 100% by CRA, will be analysed and discussed, establishing the suitability of their use.

#### **2. Materials and Methods**

#### *2.1. Materials*

In this research, the NA used were: limestone gravel (2/6, 6/12 and 12/20 mm), and silica sand (0/2 and 0/4 mm) to produce the reference concrete. For the manufacture of RAC, CRA obtained from the crushing of concrete with EAFS (using a jaw crusher) 2 months old has been used. The resulting crushing material has a range of grading of 0/25 mm. This EAFS concrete has been manufactured with cement (CEM) I 52.5 R and a w/c ratio of 0.47. This source material presents at 28 days a compressive strength of 88 MPa, a modulus of elasticity of 52 GPa and an oxygen permeability of 6.48 <sup>×</sup> <sup>10</sup>−<sup>18</sup> <sup>m</sup>2. The physical properties of both the NA and RA are shown in Table 2, after performing a measurement. The specific gravity and the water absorption have been determined according to EN 1097-6, the shape

index was obtained following the EN 933-4 and Los Angeles wear has been determined according to EN 1097-2. The Portland cement used in RAC is CEM I 42.5 R (European standard), whose density is 3.15 g/cm3 according to UNE 80103, and the mix has been made with tap water.


**Table 2.** Characterization of the aggregates.

The obtained water absorption of the CRA meets the requirements of the Spanish standard EHE-08 for structural concrete [7] and is more than twice that of coarse NA. However, this value is very small compared to more common CRA that normally exceeds 5% [5,33,57], although it depends on the size range. The shape index of the CRA is slightly lower than that of the NA and approximately one and a half times the values obtained by Etxeberría et al. [9] for conventional CRA. In terms of workability, these aggregates show a low shape index, but EAFS exposed surface is very cavernous and cause mesh between aggregates, demanding an extra volume of cement paste or mortar to fill the holes in their surface. However, the solid fraction of EAFS is generally much less absorbent than that of NA. Due to its properties, good mechanical and durability properties are expected [39].

#### *2.2. Mix Design*

The design of the mix has been made using the Faury method, to obtain maximum compactness. The maximum aggregate size has been set at 20 mm, in accordance with EHE-08. The coarse aggregates (>4 mm, EN 13139) of the reference concrete (NAC) have been replaced at several ratios (20%, 50%, and 100% vol.) with CRA. The content of cement has been set at 350 kg/m<sup>3</sup> and the effective w/c ratio of the reference concrete (NAC) at 0.5. The total w/c ratio has been determined by adding compensation water equal to that estimated for the mixing time (10 min) from the water absorption over time test, according to the method proposed by Rodrigues et al. [58]. The strength class of concrete has been defined as C30/37 in accordance with EN 1992-1-1. The slump has been defined as 70 ± 10 mm (S2) according to EN 12350-2 and without plasticizers (in order not to introduce more variables) for all replacement ratios. To maintain the same slump in all the mixes, the effective w/c has been slightly modified. RAC has been manufactured from the theoretical curve of NAC (Figure 1), maintaining between mixes the same volume of aggregates of each sieve fraction, so the mix grading for NAC and for RAC is the same. The mix proportions used are shown in Table 3.

**Figure 1.** Theoretical curve and grading of the different aggregates according to EN 933-1.


**Table 3.** Concrete mix proportions.

The aggregates were dried at 100 ± 2 ◦C until constant weight before mixing and the mixing process consisted of a sequence of 4 min with the coarse aggregates and 2/3 of the water, 2 more minutes after adding the fine aggregates and a further 4 min after adding the cement and 1/3 of the water. The concrete was demoulded after 24 h of manufacture and it has been cured in a humidity chamber at 20 ± 2 ◦C and 95 ± 2% humidity (except for drying shrinkage testing specimens).

A scheme illustrating the successive steps of RAC's manufacturing process is shown in Figure 2. The process describes the possible multi-recycling of RAC.

**Figure 2.** Recycled aggregate concrete (RAC) manufacturing process.

#### *2.3. Physical Properties Tests*

The concrete's absorption by capillarity has been determined on four cylindrical specimens with 150 mm diameter and 100 mm in length per mix proportion, after 28 days of curing in a humidity chamber and 14 days in an oven at 60 ± 5 ◦C. The test consists of measuring the mass evolution after 3, 6, 24 and 72 h of immersion, according to LNEC (National Laboratory for Civil Engineering) standards following the LNEC E-393. The absorption by immersion has been determined on four 100 mm cubic samples after 28 days of curing in a humidity chamber according to LNEC E-394. In addition, the apparent bulk, bulk, and saturated surface dry (SSD) density have been determined according to EN 12390-7 and the open porosity according to UNE 83980, for all the mixes produced.

#### *2.4. Mechanical Properties Tests*

The compressive strength (*fc*) has been obtained on 150 mm cubic samples (EN 12390-1) per mix at ages of 7, 28 and 91 days. The test specimens have been tested using a load application rate of 0.6 MPa/s in a servo-hydraulic press of 3000 kN capacity and in accordance with EN 12390-3. The ultrasonic pulse velocity test has been performed on the specimens intended for the compressive strength test, prior to testing the compressive strength and just after their surface is dry. The measurement has been carried out according to EN 12504-4 with the transducers in direct transmission placed in collinear directions between two parallel faces with Vaseline on the contact surface between transducers and the concrete surface. The pulse velocity is calculated as the *length o f specimen pulse travel time* ratio. The compressive modulus of elasticity (E) has been determined in a servo-hydraulic press of 250 kN capacity, on three cylindrical specimens with 150 mm and 300 mm in length per mix, after 28 days of curing in a humidity chamber. The upper and lower faces of the test specimens have been levelled before the test and a compressometer/extensometer equipped with high precision displacement transducers is used to measure the micro-deformation. Four loading/unloading cycles have been used, applying an initial stress of 1 MPa (17.6 kN) and a load application speed of 0.5 MPa/s (8.8 kN), using a maximum load of *fc*/3 according to LNEC E-397. The splitting tensile strength has been determined on the three specimens used in the modulus of elasticity test, for all mixes. A servo-hydraulic press of 3000 kN capacity and a load rate of 0.05 MPa/s (3.5 kN/s) was used, according to EN 12390-6.

#### *2.5. Durability Tests*

The resistance to chloride-ion penetration was determined by calculating the diffusion coefficient by means of the depth of chlorides penetration into concrete, according to LNEC E-463. Three cylindrical specimens with 100 mm diameter and 50 mm in length per mix have been used for each of the ages (28 and 91 days) and mixes. The specimens were cured in a wet chamber and moved to a dry chamber (20 ± 2 ◦C and 60 ± 5% relative humidity) in the last 14 days before testing. Carbonation resistance of the concrete was determined on three cylindrical specimens with 100 diameter and 50 mm in length per mix and per exposure time, stored 14 days in a humidity chamber followed by 14 days in a dry chamber (20 ± 2 ◦C and 60 ± 5% humidity) before being placed for 7, 28, or 91 days in the carbonation chamber. The conditions of the carbonation chamber and the test methodology are those proposed by LNEC E-391 (temperature of 23 ± 3 ◦C, relative humidity of 60 ± 5%, and CO2 concentration of 5.0 ± 0.1%). The determination of the carbonation depth was carried out with the help of a pH indicator (1% phenolphthalein solution in ethanol), cutting the specimen in quarters, spraying the solution and measuring the depth of carbonation penetration (the average depth measured in the eight contact surfaces of the broken specimen). Drying shrinkage was measured on two 100 × 100 × 500 mm prismatic specimens per mix and according to LNEC E-398, from 24 h to 91 days of age. The specimens were placed in a chamber at 20 ± 2 ◦C and 55 ± 5% relative humidity after demoulding and during the 91 days of testing.

#### **3. Results and Discussion**

#### *3.1. Physical Properties*

Table 4 shows the average physical properties of all the mixes produced and their standard deviation. All densities increase as the replacement does, since CRA has a bulk density 0.3 g/cm3 higher than that of NA (Table 2). The increase in bulk density is close to 7% for 100% replacement and opposite to that obtained by other authors using more common RCA [20–22]. This increase in density allows a saving of volume with respect to NAC in applications where self-weight is important (i.e., bridge counterweights or seawalls). To compare the amount of voids in mixes with different density, the property to be analyzed is open porosity, since the water absorption is the *dry oven mass aparent volume* ratio and hence depends on the bulk density of the material. Open porosity increases by 18% for 100% replacement, which is approximately 70% of the relative increase obtained by Thomas et al. [15] for more common RCA mixes with a w/c ratio of 0.5. The increase in open porosity with respect to the NAC is due to RA being on average 160% more porous than coarse NA (Table 2) and coarse aggregate represents 44% of RAC's volume. This large increase is due to the porous nature of the source concrete mortar. Analogously, the increase in porosity obtained is relatively low because the source concrete has a low water/cement ratio (0.47).


**Table 4.** Physical properties of hardened concrete mixes.

Figure 3 shows the capillarity absorption over time for the different replacement ratios used.

50% 2.390 ± 0.009 2.740 ± 0.01 2.520 ± 0.015 2.56 12.73 ± 0.41 5.33 ± 0.21 100% 2.460 ± 0.005 2.920 ± 0.018 2.620 ± 0.005 2.60 15.62 ± 0.53 6.34 ± 0.22

**Figure 3.** Capillarity absorption vs. time (**a**) and capillarity absorption vs. replacement ratio (**b**).

Capillarity absorption increases with immersion time (Figure 3a) and the level of replacement (Figure 3b). For 20% replacement, the variation in absorption with respect to NAC is negligible for any immersion time, while for 100% replacement the values may increase by 9% at 72 h of immersion. For replacement of 50%, intermediate capillary absorption increases of around 4% are obtained. For total replacement and 72 h immersion, other authors obtained increases of around 30% for both current RAC [19] and high-quality precast concrete RAC [10]. The higher absorption shown by RAC is due to the high porosity of RCA, which contains more and longer capillaries than NAC's. Figure 3a shows that the capillarity absorption difference is maximum at 24 h between 0% and 100% replacement and the difference is reduced at 72 h. This happens because the mix with 100% replacement has larger pores size (larger capillary pores saturate the first). The low water absorption of the study concrete is compatible with the minimum capillarity absorption obtained, which shows a loss very similar to that of ultrasonic pulse velocity, strongly related to the mechanical properties of the hardened concrete.

#### *3.2. Mechanical Properties*

There are several factors that positively and negatively affect the compressive strength of the study RAC: porosity of the cement paste, the new ITZ's quality and characteristics of the RCA. Porosity negatively affects strength. The greater porosity of RAC is due to the porosity of RCA and to that generated in the ITZ between RCA and the new paste [59]. This porosity largely depends on the w/c ratio of both the source concrete and RAC. This strength loss is not high for this RAC, since the porosity of mixes with a 100% replacement is only 18% higher than of NAC, because the source concrete has a low w/c ratio and RAC mixes have a lower effective w/c ratio than the source concrete. A lower quality ITZ has been reported by various authors [30,31] as the main cause of the strength loss of RAC, justifying that this element is the weakest link of the chain. The shape of the EAFS aggregates in RCA and the high compressive strength and stiffness of the source concrete minimize this effect. On the other hand, compared with NA, RCA's shape index is 23% lower and, even having a greater roughness, these aggregates allow using slightly lower effective w/c ratios to obtain the same workability, generating lower pores in the mortar and reducing the strength loss.

The result of these effects on the compressive strength is shown in Figure 4 where the evolution over time of the compressive strength for all mixes is shown.

**Figure 4.** Compressive strength of all mixes at different ages.

At 7 days, the compressive strength decreases slightly and linearly as replacement increases. For 100% replacement, the strength loss with respect to NAC is 5.5%. At 28 days, the concrete class for all replacement is C35/45, exceeding the C30/37 class proposed at the design phase. At 28 days, a strength loss or gain depending on the replacement cannot be clearly defined, but a relative 7-day strength increase of 21% for NAC and 23% for RAC with 100% of RA is observed. This slightly higher relative increase for high replacement could be due to a hydration of the unhydrated cement grains of the old mortar, thus improving the bond in the ITZ with the new mortar. So the curing age is an important factor to check the effectiveness of these concrete mixes. At 91 days, the replacement level does not affect compressive strength. The slope of the linear adjustment is practically 0 and the standard deviation close to 1 for all replacements, proving an increase in compressive strength with age and replacement ratio, because the old mortar is still hydrating. In any case, the compressive strength obtained is much lower than that of the source concrete, although relatively higher than that obtained in other studies for current CDW [9,17,27,32–34] which on average lose around 10% compressive strength at 28 days with respect to NAC. This is due to the high quality of the RCA used

in the present study, namely the high strength of the attached mortar of these CRA (due to the strength of the source concrete) and its low shape index.

The results of the splitting tensile strength tests are shown in Figure 5. The high standard deviation, which is typical of splitting tensile strength tests [33], and the similar mean values obtained show that the behaviour of RAC is very similar to NAC's for any replacement ratio. This happens because there is a good bond strength between RCA and the new paste, due to the higher roughness of RCA. This greater roughness seems to compensate for the weaker mortar surface generated in the crushing stage and exposed in the new ITZ. It also compensates for RAC not benefitting from the bond produced by the chemical interaction between calcium hydroxide and the calcareous aggregate [59] in NAC. As seen in Figure 6, fracture surfaces propagate through the aggregates (good bond strength in the ITZ), both for NAC and RAC. In the case of NAC, the crack surface is flatter, favored by the crystalline structure of calcite, while the fracture surface produced in CRA is irregular or lumpy with a smooth transition to the new mortar paste. Most of the macro-cracks present in RCA are produced in the mortar phase of the source concrete, even though sometimes they are also produced through the EAFS aggregate.

**Figure 5.** Tensile strength of all mixes at 28 days.

The splitting tensile strength values obtained by other authors for RAC produced with CDW are variable depending on the quality of RA. According to the classification proposed by Silva et al. [14], RA are classified in four classes according to their intrinsic properties (bulk density, water absorption and Los Angeles wear). For the most demanding category (Class A), the loss of tensile strength should be around 10% with respect to NAC [60]. The RA used in the present study, even though not complying with the water absorption requirements described for Class A, shows a tensile strength similar to NAC's.

Figure 7 shows the compressive modulus of elasticity for all mixes produced. Using a linear trend line, an inverse relationship between modulus of elasticity and replacement level can be established. While NAC shows a 40 GPa modulus of elasticity, the 100% replacement mix shows values around 33 GPa, 17% lower and with standard deviations not exceeding 0.5 MPa. This reduction is lower than that commonly reported and is related to the properties of the source concrete. The source concrete used to produce RCA has low w/c ratio, a modulus of elasticity greater than 50 GPa and a low porosity close to 10%. For similar w/c ratios and cement contents, other authors reported reductions from 15% to more than 40% [15,35,61] using RCA produced from common concrete waste.

Traditionally, density is a factor directly related to the modulus of elasticity because it is linked to the porosity of concrete [59]. The aggregate used is 7% denser than NA, but this is due more to the iron nature of RA than to the very slight increase of w/c ratio. The porosity of RCA influences its stiffness and hence the stiffness of concrete, since its increase allows greater deformations per unit of applied load.

Ultrasonic pulse velocity is intrinsically related to the modulus of elasticity and density. The ultrasonic pulse velocity obtained for all the mixes produced is shown in Figure 8. Following the trend line, it is observed that the pulse velocity decreases as the replacement ratio increases. There is a maximum loss of 6% for 100% replacement, slightly less than 9% determined by Khatib [52] for current RAC with the same w/c ratio. It should be remarked that the source concrete showed an ultrasonic pulse velocity of 5.22 km/s at the same age. On the other hand, the maximum loss obtained is similar to that obtained by de Brito et al. [10] for RAC produced from concrete precast rejects, which demonstrates the homogeneity of RAC and the high quality of the RCA used. The RCA used in this study has a density 10% and 20% higher than NA and of current RCA respectively, which a priori favours an increase in the ultrasonic pulse velocity. On the other hand, the porosity of the source concrete mortar allows RAC to strain more under sustained loads, obtaining smaller modulus of elasticity and lower ultrasonic pulse velocity than NAC, due to the discontinuities generated by the pores.

(**b**)

**Figure 6.** Crack patterns of test specimens subjected to tensile strength test for 0% (**a**) and 100% (**b**) replacement ratio.

**Figure 7.** Modulus of elasticity of all mixes at 28 days.

**Figure 8.** Ultrasonic pulse velocity at 28 days.

#### *3.3. Durability*

Carbonation is a phenomenon that alters the alkaline behaviour of concrete. When CO2 from the atmosphere diffuses into the concrete matrix, it reacts with water forming carbonic acid and the latter reacts with portlandite forming calcium carbonate. This process reduces the pH of concrete, depassivating the reinforcement. Figure 9 shows the evolution of the carbonation depth over time of all mixes. At 7 days, the carbonation depth is very small due to the low w/c ratio of the mixes. The test scatter caused by reading tolerances is relevant but mixes with 100% and 50% replacements appear to be slightly less durable. However, at 28 days, a slight increase in carbonation depth with the increase in CRA incorporation can be observed, which can reach about 10% for total replacement. This trend continues at 91 days, when carbonation becomes more noticeable for higher replacements: 50 and 100%. For total replacement, carbonation depths can be 15% higher than for reference concrete.

**Figure 9.** Carbonation depth for all mixes and ages.

The general increase in carbonation depth with replacement ratio and age is due to several causes. CRA is composed of the source concrete mortar, hence part of this cement is already carbonated. Another relevant fact is that the open porosity of RAC is 18% higher than that of NAC, which favours the increase of permeability and hence the diffusion of CO2 through the cement matrix, since there is easier access to more paste surface. On the other hand, by having a greater volume of total mortar, there is also a greater amount of total portlandite compared to the reference concrete (a greater amount of calcium carbonate can be generated) and hence a greater amount of CO2 is required for the same carbonation depth.

Other authors have also obtained increases in carbonation depth as the replacement level increases. For similar mixes and 100% replacement, de Brito et al. [10] obtained a marginal increase of the depth of carbonation for rejected high-quality precast concrete at all ages tested. Ryu et al. [62] obtained a slight increase in carbonation depth for similar mixes and similar compressive strengths, namely an increment of 14% for total replacement. Finally, in the review of Silva et al. [45], it is found that, for total replacement, the average carbonation depth may increase by a factor of 2 with respect to the control concrete, establishing the permeability of concrete, which depends directly on the water absorption (and therefore on the open porosity) of RCA, as a key factor.

Chlorides negatively affect concrete, reacting with tricalcium aluminate forming Friedel's salt (bound chlorides) and oxidizing the reinforcement after penetrating concrete (free chlorides) [63]. The resistance to the penetration of chlorides is shown in Figure 10. The chlorides diffusion coefficient increases with replacement ratio and decreases with age. At 28 days, there is a quasi linear increase in this coefficient, reaching 11% for total replacement. At 91 days, the increase is 9.5% for total replacement, due to the extra contribution of mortar from the source concrete that continues to hydrate. On the other hand, there is a large reduction in chlorides diffusion coefficient with age, a reduction greater than 25% for all replacements.

**Figure 10.** Chloride diffusion coefficient for all mixes and ages.

Chlorides penetration in concrete (an advective-diffusive phenomenon) largely depends on its porosity and the tortuosity of its capillary framework [64]. A strong correlation has been obtained between the chlorides diffusion coefficient and both open porosity (R<sup>2</sup> = 0.73) and capillarity absorption (R2 = 0.89). Critical variables such as the content or type of cement remain constant, so it seems consistent that the higher ingress of chlorides in RAC with larger replacement is due to the increase in open porosity, owing to the higher volume of voids in RCA.

The incorporation of EAFS usually improves the chloride diffusion coefficient relative to the reference concrete by about 10% [56], so it is expected that the behaviour of RAC with RCA based on EAFS can show an improvement over mixes with current RCA. Likewise, concrete with current RCA generally have higher chloride diffusion coefficient than concrete produced with NA. Bravo et al. [24] obtained average increases of more than 40% in the chlorides diffusion coefficient for total replacement of CRA produced from unsorted CDW. Other studies, however, argue that this property is unrelated with the RA replacement level [10], stating that it depends mostly on the quality of CRA.

Drying shrinkage is a significant phenomenon during cement hydration. The difference in relative humidity between the environment and the specimen causes the water adsorbed to the hydrated paste to pass into the atmosphere following Fick's law. Drying the specimen involves a reduction in volume that can cause cracking of the paste in severe cases. The evolution of drying shrinkage for the different mixes is shown in Figure 11. The drying shrinkage obtained for RAC with total replacement is about 530 μm/m at 91 days, while for NAC it is 457 μm/m, so there is an increase of 13%.

**Figure 11.** Mean drying shrinkage over time for all mixes.

It is well known that there is a strong relationship between drying shrinkage and modulus of elasticity in concrete [65]. The shape and texture of aggregates are variables that affect the drying shrinkage, although modulus of elasticity (or any of the variables that affect the latter) is the parameter that most affects the former [59]. An increase in modulus of elasticity generates a greater restriction against strain, i.e., stiffer aggregates more efficiently oppose the strain generated by shrinkage than more deformable ones. As stated before, the modulus of elasticity drops with increase of the replacement ratio (17% for full replacement) and drying shrinkage decreases similarly. Another factor to keep in mind is that RCA is still shrinking, especially if the source concrete is young concrete, while NA do not shrink. Recycling concrete that has already reached the end of its service life is not a problem, but in the case of laboratory-produced concrete (the case here) it is something to keep in mind. Shrinkage measured from RAC with laboratory-produced source concrete is thus expected to overestimate the shrinkage of RAC. Losses for total RCA replacement range from 10% [10] to 50% [43], the latter being the value established in the Xuping review as a typical value for RAC.

#### **4. Conclusions**

In this study, a new concrete based on the use of concrete waste has been tested. The concrete waste used incorporates a by-product of the steelmaking industry (EAFS), thus waste production and natural aggregate consumption are reduced. The physical, mechanical and durability properties of the new material have been characterized and compared with the results of a reference conventional concrete with the same mix design of the recycled aggregate concrete tested. Moreover, experiments from other authors on the properties of concrete with recycled aggregates from other sources were also compared with those of this experiment. The following conclusions were drawn:

• The use of RCA with a water absorption twice that of NA requires compensation water, but the rounded shape of the RCA produced from concrete with EAFS (shape index 20% lower than NAC) makes it possible to use slightly lower effective w/c ratio to obtain the same workability;


**Author Contributions:** Data curation, J.d.B.; Formal analysis, P.T., J.P., C.T. and J.d.B.; Funding acquisition, C.T. and J.R.; Investigation, P.T., J.P. and C.T.; Methodology, C.T.; Project administration, C.T.; Resources, J.R.; Writing-original draft, P.T. and J.P.; Writing-review & editing, C.T. and J.d.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was co-financed by the European Regional Development Fund (ERDF) and the Ministry of Economy, Industry and Competitiveness (MINECO) within the framework of the project RTC-2016-5637-3 and in collaboration with the company INGECID and the department LADICIM (University of Cantabria).

**Acknowledgments:** The authors gratefully acknowledge the support of the CERIS Research Unit, IST, University of Lisbon and of FCT.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **High-Frequency Fatigue Testing of Recycled Aggregate Concrete**

#### **Jose Sainz-Aja, Carlos Thomas \*, Juan A. Polanco and Isidro Carrascal**

LADICIM (Laboratory of Materials Science and Engineering), University of Cantabria, E.T.S. de Ingenieros de Caminos, Canales y Puertos, Av./Los Castros 44, 39005 Santander, Spain; jose.sainz-aja@unican.es (J.S.-A.); polancoa@unican.es (J.A.P.); carrasci@unican.es (I.C.) **\*** Correspondence: carlos.thomas@unican.es

Received: 11 November 2019; Accepted: 11 December 2019; Published: 18 December 2019

**Abstract:** Concrete fatigue behaviour has not been extensively studied, in part because of the difficulty and cost. Some concrete elements subjected to this type of load include the railway superstructure of sleepers or slab track, bridges for both road and rail traffic and the foundations of wind turbine towers or offshore structures. In order to address fatigue problems, a methodology was proposed that reduces the lengthy testing time and high cost by increasing the test frequency up to the resonance frequency of the set formed by the specimen and the test machine. After comparing this test method with conventional frequency tests, it was found that tests performed at a high frequency (90 ± 5 Hz) were more conservative than those performed at a moderate frequency (10 Hz); this effect was magnified in those concretes with recycled aggregates coming from crushed concrete (RC-S). In addition, it was found that the resonance frequency of the specimen–test machine set was a parameter capable of identifying whether the specimen was close to failure.

**Keywords:** high-frequency fatigue test; recycled aggregate; recycled aggregate concrete; fatigue; Locati test

#### **1. Introduction**

Recycled aggregates have an environmental benefit [1,2] and recycling is a necessity as the rate of waste generation is such that landfills are close to saturation [3]. Additionally, the possibility of obtaining recycled concrete (RAC) with good mechanical [4] and durability [5] properties has been proven. The ability of these recycled aggregates for use in self-compacting concrete [6], even with fine recycled aggregates [7], has also been proven. Even Kareem et al. [8] have used RAC for the manufacture of hot-mix asphalt.

Concrete fatigue behaviour has not been extensively studied, partly because of the difficulty, cost and time required. Concrete elements subjected to this type of load include railway superstructures, sleepers or slab tracks [9], rail and road bridges [10], offshore structures subject to variable wind and tidal loads [10,11] and/or wind generators [12]. Khosravani et al. [13] have defined a procedure for analyzing ultra-high performance concrete's response to impacts. As Skar ˙zy ´nski et al. state, knowledge about the effect of cyclic loads on concrete is currently very limited [12]. Several authors have analysed the responses of concrete. Xiao et al. [11] analysed the behavior of RAC to both compression and bending fatigue. Li et al. [14] analysed the influence of compressive fatigue on a fiber-reinforced cementitious material. Thomas et al. analysed the concrete fatigue behavior using two different methods: the staircase method [15] and the Locati method [16]. Innovative techniques, such as micro computed tomography (micro-CT) have also been used to analyze the behavior of concrete at fatigue, both in compressive fatigue [12] and bending fatigue [17]. Moreover, some micro-CT studies, which is a technique able to analyse the pores and cracks on concrete [18,19], had been developed to understand the concrete damage fatigue micromechanism [20,21]. In general, fatigue is known to lead

to microcracks in concrete growing at lower loads than in static tests, which can lead to concrete failure earlier than expected [22].

It is well known that the fatigue limit of concrete depends on different factors. At higher stresses, the fatigue strength decreases with decreasing frequency [12,23,24]. The fatigue strength is also affected by the water/cement ratio, cement content, concrete type, rest periods, curing conditions and age during loading [25]. It is assumed that damage linearly increases with the number of cycles applied at a certain stress level [24]. The strain at the concrete failure during fatigue tests approximately corresponds to that at the peak load during quasi-static tests [26]. The failure meso-mechanism in concrete under fatigue compressive tests is almost the same as in monotonic compressive tests [27].

In order to reduce the characterization time as much as possible, the influence of increasing the test frequency up to the resonance frequency of the set of specimens and test machine [28] has been analyzed. In response to this proposal, several authors [23,29,30] indicate that the range of frequencies to be tested should be between 1 and 15 Hz, since they state that within this range, the effect of frequency is limited. It is justifiable to set the minimum at 1 Hz in order to avoid an excessive increase in the effect of creep in very long duration tests. However, there is no justification for setting the maximum at 15 Hz.

Fatigue tests are proposed for classification according to the frequency of the test, distinguishing between low frequency tests, moderate frequency tests and high frequency tests. Low frequency tests are carried out at less than 1 Hz, moderate frequency tests between 1 and 15 Hz and high frequency tests at more than 15 Hz.

Three types of recycled self-compacting concrete were characterized in this investigation of compression fatigue tests. This characterization was developed both at a moderate frequency (10 Hz) and a high frequency (90 Hz). First, the results of the Locati tests were compared with 2 <sup>×</sup> 105 cycles per step at moderate and high frequencies, where it was possible to show that the tests carried out at a high frequency were notably more conservative than those at a moderate frequency. Second, the influence of the number of cycles per step in the high frequency Locati tests was analysed, where similar results were obtained using both test methodologies.

#### **2. Materials and Methods**

#### *2.1. Aggregates*

Recycled aggregates used for the manufacture of concretes were obtained from the crushing of ballast (RA-B) and out-of-use sleepers (RA-S). After crushing these materials, three granulometric fractions were obtained from each of these wastes (see Table 1). The grading of the six fractions of the two crushed products can be seen in Figure 1. Additionally, Table 1 shows the results of the relative density of the coarse aggregate and the real densities of the sands. The flakiness index of the two coarse aggregates was also determined, obtaining a value of 14% for RA-B-CA and 5% for RA-S-CA.



**Figure 1.** Aggregate grading curves.

#### *2.2. Cement*

A CEM IV (V) 32.5 N type cement according to EN 197-1 [31] was used, provided by Alpha cements [32], with a density of 2.85 g/cm3 determined according to UNE 80103 [33] and a Blaine specific surface of 3885 cm2/g obtained according to EN 196-6 [34]. The chemical composition of the cement is given in Table 2.

**Table 2.** Cement chemical composition.


#### *2.3. Mix Proportions*

Three self-compacting recycled concretes were mixed with recycled aggregates from out-of-use track elements: the first, exclusively with RA-B, called RC-B; the second, exclusively with RA-S, called RC-S; and finally, a third concrete with both types of aggregates in the proportion in which these wastes were found in a ballast track, called RC-M. Table 3 shows the three mix proportions.

**Table 3.** Mix proportions.


It should benoted that themixproportionsweredeterminedusing the criteria that the three self-compacting concretes had a similar workability; therefore, as the RA-B-CA had a notably sharper geometry than the RA-S-CA, it was necessary to increase the amount of water to achieve the same workability [35].

#### *2.4. Mechanical Properties*

The compressive strength tests were performed according to the EN 12390-3 and EN 13290-3/AC [36,37] standards, using cubes of 100 mm side at ages of 1, 2, 3, 5, 7, 28, 90 and 180 days, testing 5 specimens for each age and material. The stabilized secant moduli of elasticity was determined according to EN 12390-13 [38] using cylindrical specimens of 200 mm in height and 100 mm in diameter at the ages of 7, 28, 90 and 180 days, testing 2 specimens for each age and material.

#### *2.5. Fatigue Tests*

The influence of conducting trials at a very high frequency was analyzed. For this purpose, the reference was tests carried out at moderate frequency, i.e., usual test frequencies, namely 10 Hz. For very high frequency tests, the highest possible test frequency was used, namely, tests were performed on a resonant fatigue machine. This machine carried out the tests at the resonance frequency of the set test machine and the specimen. In the specific case of the resonance machine used and the specimens used, these resonance frequencies were in the range of 90 ± 5 Hz.

An explicative scheme of the fatigue tests that were carried out can be found in Figure 2a. Fatigue tests at both a moderate frequency (10 Hz) and a very high frequency (90 Hz) were performed using cylindrical specimens of 200 mm in height and 100 mm in diameter in all tests when these specimens were older than 90 days; this was done to ensure that the properties of the concrete had reached a stationary state. All the tests carried out recorded both the load applied to the specimen and the evolution of the strain suffered by the specimen. The strain was recorded by means of two strain gauges fixed to the specimens in two diametrically opposed generatrixes by considering the mean of the values provided by both gauges as the strain value of the specimen. In the high-frequency tests, the frequency at which the system was located was recorded, noting that the resonant frequency of the system will evolve with the variations in its stiffness. For each of the fatigue test types, two specimens were tested, ensuring that similar results were obtained in both.

**Figure 2.** (**a**) Experimental fatigue campaign. (**b**) Locati methodology description. MF: moderate frequency, HF: high frequency.

Fatigue characterization at a moderate frequency was performed using the Locati method [39,40]. The Locati methodology consists of applying increasing steps of sinusoidal loads, maintaining a constant ratio <sup>σ</sup>max/σmin <sup>=</sup> 0.1 over a constant number of cycles, in this case 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles (Figure 2b). These tests were performed on a servo-hydraulic machine with a maximum capacity of 1000 kN (Figure 2b). The criterion used to determine the fatigue limit using this type of test was that followed by Thomas in his PhD thesis [41], which obtained the fatigue limit stress range to be 80% of the stress range of the

step in which the breaking occurs. The stress range was the difference between the maximum and minimum stress, i.e., Δσ = σmax − σmin. This method is called method-1.

In order to analyse whether an increase in frequency had an influence on the concrete fatigue limit, tests were performed under identical loads at a moderate frequency and a very high frequency. A 400-kN capacity machine was used to perform fatigue tests at the resonance frequency of the test set. This resonance frequency was found in all cases to be in the range 90 ± 5 Hz (see Figure 3b). In this case, in addition to analysing the results following the procedure followed by Thomas in his PhD thesis [41], an additional failure criterion was introduced, namely, to define the stress range of the step prior to a resonance frequency drop as the stress range of the endurance limit. This method was denominated as method-2.

**Figure 3.** Fatigue tests. (**a**) Low-frequency fatigue testing and (**b**) high-frequency fatigue testing.

In order to analyse the influence of the number of cycles in the Locati tests, very high frequency tests were performed with the same stress values but applying 5 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step instead of 2 <sup>×</sup> 105 cycles. For these tests, the same testing machine was used, and the same criteria were used to define the endurance tension range.

The common objective of all these studies was to determine the step associated with the fatigue limit of the concrete in each case. In order to be able to compare the results as directly as possible, it was decided to fix the load steps based on the compressive strength of each material. For this purpose, the value of the maximum stress of each step was defined as a coefficient (k) multiplied by the compressive strength at the time the fatigue tests began, that is, 90 days after manufacture of each of the concretes; furthermore, the tensional ratio σmax/σmin was set to 0.1. Table 4 shows a summary of the tension values used in each of the load steps applied.


**Table 4.** Fatigue test stress values.

In order to compare the degree of influence of repeated loads on each of the materials, the influence coefficient (IC) was defined as the ratio of the stress range corresponding to the fatigue limit of the concrete to its compressive strength.

#### **3. Results and Discussions**

#### *3.1. Compressive Strength and Young's Modulus*

Figure 4 shows the evolution of the compressive strength as a function of time, while Figure 5 shows the evolution of Young's modulus as a function of the age of the different concretes.

**Figure 5.** Evolution of the Young's modulus.

The great influence of the water/cement ratio on the compressive strength of the RC-S can be observed. On the other hand, RC-B had the lowest compressive strengths and RC-M had intermediate compressive strengths between the other two concretes.

In the case of Young's modulus, although the RC-S paste was of a better quality than that of RC-B, the noticeably lower stiffness of the mortar, adhered to the natural aggregates that made up the RA-S, meant that the elastic modulus of RC-S was lower than that of RC-B. As in the case of the compressive strength, RC-M was found between RC-B and RC-S concretes in all cases.

#### *3.2. Influence of the Frequency on Fatigue*

Figure 6 shows an example of the maximum deformation envelope recorded during the Locati tests at a frequency of 10 Hz.

**Figure 6.** Maximum strain of low-frequency Locati fatigue tests with 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step.

Figure 6 shows, first of all, that RC-B was able to resist the most steps, which meant that it was the material with the highest IC, with RC-S being able to resist the least and RC-M was in an intermediate situation between the other two materials. These results agree with the results of other authors who state that the presence of adhered mortar in the RA reduces this coefficient [15,16,42]. It can also be seen that the deformation values suffered by the specimens was lower in the case of RC-B. It should be noted that, as each of the steps is a fixed percentage compared to the compressive strength of each material, they are not directly comparable to each other as they are different stress values.

Table 5 shows a brief summary of the results obtained from the moderate-frequency tests carried out using method-1.


**Table 5.** Low-frequency fatigue limit (fL) obtained using the Locati method with 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step. IC: influence coefficient.

Analysis of these results shows that the material with the highest stress range was RC-S, the least was RC-B and the RC-M had an intermediate value. Likewise, it is possible to determine that, although the RC-S had the highest stress range values, it had the lowest IC. This loss of compressive strength was due to the presence of mortar adhered to the aggregate. In any case, these values of IC are within the usual range for concretes and agree with values found in the literature [15,16].

Figure 7 shows an example of the maximum deformation suffered by a specimen of each material during Locati tests at the resonance frequency of the machine test set.

**Figure 7.** Maximum strain of high-frequency Locati fatigue tests with 2 <sup>×</sup> 105 cycles per step.

Figure 7 shows that RC-S was the material that had the lowest IC. RC-M was generally between RC-B and RC-S.

Figure 8 shows that the resonance frequency evolved throughout the test. For the first steps, there was an increase in the frequency with the cycles throughout each step, as well as a punctual increase when there was a change of step. At the end of each test, a drop in the resonance frequency of the system was seen. This resonance frequency depended on the stiffness of the system; an increase in the stiffness of the system resulted in an increase in the resonance frequency, while a reduction in the stiffness of the system reduced it. For this reason, it was interpreted that both point and distributed frequency increases occurred as a consequence of a stiffening of the system, while the fall that occurred in the final part of the test was a consequence of the damage suffered by the specimen, where the cracks had reached such a size that they produced a flexibilization of the specimen, which indicated that it was close to breaking.

**Figure 8.** Maximum strain and resonance frequency comparison during high-frequency Locati fatigue tests with 2 <sup>×</sup> 105 cycles per step.

Table <sup>6</sup> shows the results obtained from the Locati tests with 2 <sup>×</sup> 105 cycles per step at a very high frequency using the two criteria previously established.


**Table 6.** High-frequency fatigue limit obtained using the Locati method with 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step.

Analysing the results of Table 6, it can be determined that, as had been deduced from the evolution of the maximum strain, RC-B was the material with the highest IC, although the compression strength of RC-S was the highest of all. The stress ranges corresponding to the fatigue limit of the three concretes were similar. It can also be appreciated that, although the values of both the fatigue limit and IC, obtained using the two analysis criteria were similar, a 5% difference was found relative to the limit provided by Thomas [41], which was usually more conservative than the one obtained using the resonance frequency.

#### *3.3. Influence of the Number of Cycles Per Step during a Locati Test*

In order to determine the influence of increasing the number of cycles in each step of the Locati test, identical tests were carried out to those described above, but the number of cycles per step was set at 5 <sup>×</sup> <sup>10</sup><sup>5</sup> instead of 2 <sup>×</sup> <sup>10</sup>5. Figure <sup>9</sup> shows the evolution of the maximum deformation as a function of the number of cycles throughout the Locati test for 5 <sup>×</sup> 105 cycles per step.

**Figure 9.** Maximum strain of high-frequency Locati fatigue tests with 5 <sup>×</sup> 105 cycles per step.

Figure 9 shows that RC-S was the material that had the lowest IC. RC-M was between RC-B and RC-S. In this case, as in the previous case, the two analysis criteria were used to determine the stress range corresponding to the fatigue limit. These results are shown in Table 7.


**Table 7.** High-frequency fatigue limit obtained using the Locati method with 5 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step.

In order to compare the evolution of the maximum strain throughout the test, it was decided to divide the number of cycles performed by the number of cycles per step, in this way it was be possible to compare the results of the three variants of the test. An example of the comparison between the Locati test types for each material is given in Figure 10; Figure 12.

Figure 10 shows the influence of increasing the frequency on the fatigue behaviour of the RC-B by comparing the RC-B-HF-2 <sup>×</sup> 105 with the RC-B-MF-2 <sup>×</sup> 105. It is possible to conclude that, in the first phase, increasing the test frequency had no influence on the deformation suffered by the specimens. After step 4, the two curves separated, increasing more rapidly in the case of RC-B-HF-2 <sup>×</sup> 105. This behaviour can be justified by arguing that, in the first phase, a phase in which the concrete was not damaged, the effect of the frequency seemed irrelevant, whereas when the cracks reached a critical size, an increase in frequency produced a reduction in the number of cycles that RC-B could withstand. It was observed that the specimens tested at very high frequencies showed a markedly higher temperature increase than in the case of tests performed at moderate frequency. This fact is believed to be related to the reduction in fatigue life of the material.

**Figure 10.** Comparison of the maximum strain of the RC-B for a Locati test with 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step at a low frequency (RC-B-MF-2 <sup>×</sup> 105), a Locati test with 2 <sup>×</sup> 10<sup>5</sup> cycles per step at a high frequency (RC-B-HF-2 <sup>×</sup> <sup>10</sup>5) and a Locati test with 5 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step at a high frequency (RC-B-HF-5 <sup>×</sup> <sup>10</sup>5).

From this same Figure 10, it is also possible to analyse the influence of increasing the number of cycles per step from 2 <sup>×</sup> 105 to 5 <sup>×</sup> 105. Comparing the two curves, it is observed that the effect of each step was similar until the specimen was close to breaking. For this reason, it can be assumed that the influence of increasing the number of cycles per step during a Locati test was small given that the difference between steps had a greater influence than increasing the number of cycles per step by 150%.

Figure 11 shows the influence of increasing the frequency on the fatigue behaviour of the RC-S by comparing RC-S-HF-2 <sup>×</sup> 105 with RC-S-MF-2 <sup>×</sup> 105. As in RC-B, it can be concluded that, in the first phase, increasing the test frequency had no impact on the deformation suffered by the specimens. From step 2, it can be seen how the two curves split, increasing more rapidly in the case of the RC-S-HF-2 <sup>×</sup> 105. This effect was more severe in RC-S than in RC-B, which was reflected by a greater variation in the IC of RC-S as the test frequency increased.

Figure <sup>11</sup> shows the influence of increasing the number of cycles per step from 2 <sup>×</sup> 105 to 5 <sup>×</sup> <sup>10</sup><sup>5</sup> for RC-S. A similar behaviour to that of RC-B was observed, but in this case, the effect of increasing the number of cycles per step was greater than in RC-B, which was reflected in the accelerated separation of the curves. As in the case of RC-B, in this case, the greatest difference in fatigue limit was found between RC-B-HF-2 <sup>×</sup> 105 and RC-B-HF-5 <sup>×</sup> 10<sup>5</sup> and was one step; for this reason, it can be assumed that the influence of increasing the number of cycles per step during a Locati test was small, given that the jump between steps had a greater influence than increasing the number of cycles per step by 150%.

**Figure 11.** Comparison of the maximum strain of the RC-S for a Locati test with 2 <sup>×</sup> 105 cycles per step at a low frequency (RC-S-MF-2 <sup>×</sup> 105), a Locati test with 2 <sup>×</sup> 10<sup>5</sup> cycles per step at a high frequency (RC-S-HF-2 <sup>×</sup> <sup>10</sup>5) and a Locati test with 5 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step at high frequency (RC-S-HF-5 <sup>×</sup> 105).

In Figure 12, both the frequency and number of cycles affected the fatigue life during the Locati RC-M test. The behaviour of this material was an intermediate situation between RC-B and RC-S, being in all cases, more similar to the behaviour of RC-B.

**Figure 12.** Comparison of the maximum strain of the RC-M for a Locati test with 2 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step at a low frequency (RC-M-MF-2 <sup>×</sup> 105), a Locati test with 2 <sup>×</sup> 10<sup>5</sup> cycles per step at a high frequency (RC-M-HF-2 <sup>×</sup> <sup>10</sup>5) and a Locati test with 5 <sup>×</sup> <sup>10</sup><sup>5</sup> cycles per step at a high frequency (RC-M-HF-5 <sup>×</sup> <sup>10</sup>5).

To summarise, it can be stated that performing tests at a very high frequency reduced the number of cycles that a specimen was capable of resisting compared to performing the test at a moderate frequency. On the other hand, it was also demonstrated that 2 <sup>×</sup> 105 cycles per step was enough to characterize a fatigued concrete.

#### *3.4. Comparison of the Di*ff*erent Methodologies*

In this section the endurance values for each of the three concretes tested was compared using the five proposed analysis criteria. Figure 13 shows the values of the stress range corresponding to the fatigue limit of the three concretes analysed using the five previously defined procedures.

**Figure 13.** Comparison of the high-frequency fatigue limit obtained using five different procedures and the low-frequency fatigue limit.

It can be observed that, in all cases, the fatigue limits obtained using moderate frequency tests gave higher values than in very high frequency tests, regardless of the characterization method.

On the other hand, the two methods used to determine the fatigue limit at a very high frequency produced similar values in all cases, where the one that defines the fatigue limit as 80% of the tension range of the breaking step being more conservative in all cases. This was due to the weakness introduced by the mortar adhered to the aggregates, as well as RC-S being the material most affected by creep.

#### **4. Conclusions**

Although concrete is a material that is not usually characterized under fatigue, there is no doubt that there are certain elements typically made from concrete that are subjected to variable loads, and therefore, which need to be characterized under fatigue. In this work, three types of concrete were characterised under fatigue using the Locati accelerated fatigue method. Fatigue tests were carried out to analyse two fundamental variables in the duration of the tests, namely the frequency of testing and the number of cycles per step of the Locati method. The following conclusions can be observed from the results obtained:

• In all cases, the material with the lowest fatigue limit/compression resistance ratio was RC-S, which was due to the weakness introduced by the mortar adhered to the aggregates.


**Author Contributions:** J.S.-A.: investigation, methodology, formal analysis, writing original draft. C.T.: formal analysis, surpervision, writing original draft, review and editing. I.C.: founding, methodology, formal analysis, surpervision, review and editing. J.A.P.: methodology, formal analysis, surpervision, review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded Ministerio de Economía y Competitividad grant number MAT2014-57544-R.

**Acknowledgments:** The authors would like to thank The LADICIM, Laboratory of Materials Science and Engineering of the University of Cantabria and Instituto Superior Técnico of the University of Lisbon for making the facilities used in this research available to the authors.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *Appl. Sci.* **2020**, *10*, 10


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Applied Sciences* Editorial Office E-mail: applsci@mdpi.com www.mdpi.com/journal/applsci

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18