4.1. Compressive Strength
To analyze if the segregation of concrete directly affects the compressive strength, an ANOVA analysis complemented by the Tukey’s test (Honestly-significant-difference—HSD) has been carried out qualitatively evaluating this parameter. The segregation index, whose values are between 12% and 49%, were contrasted with the values of average compressive strength—average between the four cores extracted from each half ()—of each sample. To analyze qualitatively, the data were grouped into categories: 10% ≥ SI > 20%; 20% ≥ SI >30%; 30% ≥ SI > 40% and 40% ≥ SI > 50%. In total 160 observations were made (two halves of 80 samples).
Given the R
2, 10% of the variability of the dependent variable
is explained by the explanatory variable (SI). Given the p-value of the F statistic computed in the ANOVA table (
Table 3), and given the significance level of 5%, the information brought by the explanatory variable (SI) is significantly better than what a basic mean would bring, indicating that parameter SI affects the average compressive strength.
The Tukey’s test results have classified the four categories into two groups (A and B), indicating the existence of significant differences between them. As seen in
Table 4, the behavior of the average compressive strength as a function of segregation seems to follow two different patterns, one for values below the category 20% to 30% and another to values above the category 20% to 30%.
Figure 3. exemplifies two situations where a series of LWAC was vibrated in two layers (
Figure 3a) and in one layer (
Figure 3b). The graphics of all the series (16 in total) were included in the
Supplementary Data.
Figure 4a represents the variation of
as function of the segregation index. Results are consistent with the results of the ANOVA and the Tukey’s test. Concretes that have segregation indexes lower than 30% do not present any correlation between
and SI. However, starting at 30%, as SI increases,
increases. It probably occurs because at high values of SI the samples are clearly divided in two phases (mortar in the bottom and aggregates in the top), resulting in higher compressive strengths in those sections with mortar concentration. This analysis becomes more evident when the data are represented by concrete type (
Figure 5a) and it is verified that the correlations of LWAC1 and LWAC2, which has presented SI lower than 30%, are not good: 0.07 and 0.034, respectively.
Concretes vibrated in two layers (LWAC1 and LWAC2) have presented less segregation, for vibration times even higher than the concretes vibrated in one layer, and because of their better homogeneity presented minor variations in the compressive strength of its sections. These results were also discussed in previous works [
26].
From a safety point of view, the decrease in strength (sections that would contain an excessive amount of aggregates due to segregation) is more relevant than the increase in strength (sections that would contain an excessive amount of mortar), once those are mostly the areas where the concrete failure begins [
35].
Figure 4b represents the variation of
as function of the SI.
The results are also consistent with ANOVA and Tukey’s test results. Concretes with SI lower than 25% do not present any correlation between
and SI. However, starting at 25%, as SI increases,
increases, probably because at high values of SI, the samples are clearly divided in two phases (mortar in the bottom and aggregates in the top), resulting in lower compressive strengths in those sections with aggregate concentration. In the same way, this analysis becomes more evident when the data are represented by concrete type (
Figure 5b) and it is verified that the correlations of LWAC1 and LWAC2, which has presented SIs lower than 25%, are very poor: 0.00 and 0.02, respectively. The volumetric fraction of aggregates—related to the theoretical densities in this paper—also affected the reduction of compressive strength due segregation. Concretes manufactured in one layer, and with density of 1700 kg/m
3 presented higher reductions of compressive strength due segregation than concretes with theoretical densities of 1900 kg/m
3 (one layer).
4.3. Structural Efficiency
In general, the compaction in two layers led to a reduction in the overall density of the samples, most likely because the trapped air was removed more effectively in these cases. This can be seen when comparing the graphs in
Figure 7a,
Figure 8a.
Two-layer compaction also affected the compressive strength results. When comparing the graph in
Figure 7a (vibration in one layer) with the graph in
Figure 8a (vibration in two layers), it can be seen that all the curves moved to the right, so that the concretes vibrated in two layers presented superior compressive strength results.
Although the trends and the results of R
2 vary depending on the position of each eighth within the samples, in both one-layer and two-layer vibrating concretes, it can be seen through
Figure 7 and
Figure 8 that the denser, the greater in terms of compressive strength.
Concretes compacted in one layer have been shown to be more sensitive to vibration than concretes vibrated in two layers, especially in the upper region of the samples. When comparing
Figure 7b with
Figure 8b, and
Figure 7c with
Figure 8c, it can be seen that the curves of
Figure 7b,c show more vertical trends than the curves of 8b and 8c, indicating that the concretes vibrated in one layer showed a greater loss of resistance of its upper sections due to the increase in the concentration of lightweight aggregates due to segregation, which resulted in a consequent reduction in density.
To analyze if the segregation of concrete directly affects structural efficiency, the data were also analyzed using an ANOVA complemented by a Tukey test (HSD). The segregation results were contrasted with the values of average structural efficiency—average between the results of four cores extracted from each half (—of each sample. In the same way that the compressive strength analysis was performed, the data were grouped into categories: 10% ≥ S > 20%; 20% ≥ SI > 30%; 30% ≥ SI > 40% and 40% ≥ SI > 50%. In total, 160 observations were made (two halves of 80 samples).
Given the R
2, 10% of the variability of the dependent variable
is explained by the explanatory variable (SI). Given the p-value of the F statistic computed in the ANOVA table (
Table 5), and given the significance level of 5%, the information brought by the explanatory variable (SI) is significantly better than what a basic mean would bring, indicating that parameter SI average structural efficiency.
The Tukey’s test results have also classified the four categories into two groups (A and B), indicating the existence of significant differences between them. As seen in
Table 6, the behavior of the average structural efficiency as a function of segregation seems to follow two different patterns, one for values below the category 20% to 30% and another to values above the category 20% to 30%.
Figure 9a represents the variation of
as function of the segregation index. Results are consistent with the results of the ANOVA and the Tukey’s test. Concretes that have segregation indexes lower than 30% do not present any correlation between
and SI. However, starting at 30%, as SI increases,
increases. This analysis becomes more evident when the data are represented by concrete type (
Figure 10a) and it is verified that the correlations of LWAC1 and LWAC2, which has presented SI lower than 30%, are not good: 0.01 and 0.00, respectively.
Like the results obtained by Solak et al. [
26], concretes vibrated in two layers (LWAC1 and LWAC2) have presented less segregation, for vibration times even higher than the concretes vibrated in one layer, and because of their better homogeneity presented minor variations in the compressive strength, density, and as result, minor variations of structural efficiency on its sections.
Figure 9b represents the variation of
as function of the SI. Concretes with SI lower than 30% do not present any correlation between
and SI. However, starting at 30%, as SI increases,
increases. Similar to the previously analyzed parameter, this analysis becomes more evident when the data are represented by concrete type (
Figure 10b) and it is verified that the correlations of LWAC1 and LWAC2, which presented SIs lower than 30%, are very poor: 0.0006 and 0.0404, respectively.
For all structural efficiency parameters analysed (
,
and
), the concretes that were vibrated in two layers (LWAC 1 and LWAC2, green and blue in the
Figure 11), presented higher values when compared to the concretes that were vibrated in one layer (LWAC 3 and LWAC 4, yellow and red in the
Figure 11). It is not possible to state that the segregation caused significant impacts on the structural efficiency of the materials, since no clear trends were found for the four studied concretes. This is verified through the linear and approximately horizontal aspects of the data distributions, as well as through the low values of R
2.