The Effect of Surface Properties of Steel Sections on Bond Strength in Soil-Cement Mix

Abstract

Soil strengthening with hydraulic binders has gained popularity in recent years and provides an alternative to traditional methods, both for foundation reinforcement and for retaining walls. In many cases, columns, walls, or soil-cement mix blocks require reinforcement with steel sections. Correctly assessing the load-bearing capacity of a reinforced element requires an understanding of the bonding forces between the steel and the soil-cement mix. This article presents the results of pull-out tests conducted on steel flat bars embedded in a soil-cement mix. A soil-cement mix containing sand, silt, and clay fractions was prepared. The surfaces of the flat bars were treated in three different ways, and their roughness was subsequently measured. The pull-out strength of steel flat bars embedded in a soil-cement mix with compressive strength in the range of 1–2 MPa was determined. The tests revealed a correlation between surface roughness and bond strength. The conducted tests provided the basis for developing new research directions and for formulating a new bonding model for the interaction between steel profiles and soil-cement.

1. Introduction

In situ mixing of soil with hydraulic binders is a well-documented and widely applied set of techniques used in geotechnical engineering [1]. One of the methods used in wet deep soil mixing is the Continuous Deep Mixing Method (CDMM) [2], which is an alternative to DSM column technology [3]. In the CDMM, the mixing tool consists of special chain-mounted blades. Mixing occurs along the entire depth of the inserted blades, and the high mixing energy produces a homogeneous soil-cement mass with consistent strength parameters. As the machine equipped with the rotating chain advances, it forms a continuous soil-cement panel in the ground. For the foundations of engineering structures, intersecting panels can be used to create spatial grid systems [4,5]. This type of ground improvement assumes no connection between the abutment and the soil-cement panels; however, this approach is limited when excessive horizontal forces are present. One solution is to reinforce the panels with steel elements, which act as connectors between the soil-cement mix and the support structure. An alternative approach is to use prefabricated piles inserted into the soil-cement mix; this is the basis of the recently developed pre-bored grouting planted (PGP) pile technology [6,7,8]. Continuous soil-cement panels with reinforcement are also used as retaining structures, where the common approach is to consider the steel as the sole load-bearing element, while the soil-cement mix serves as filler and as the medium transferring ground pressure to the steel beams [9,10]. Whether in ground improvement applications or when panels are used as retaining walls, information about the bond strength between steel and soil-cement mix may enhance the design of such structures.

In reinforced concrete structures, the bond between reinforcing bars and concrete has been extensively described in the literature [11,12,13], and bonding models have been incorporated into design standards [14]. Research now also addresses detailed issues such as the influence of modern mixtures and concreting techniques on bonding conditions [15,16]. The bond of steel elements, whether in the form of reinforcing bars or steel profiles, remains a subject of scientific investigation in soil-cement mixes. Lepakshi and Venkatarama Reddy [17] investigated the bond forces between reinforcing bars and cement-stabilized rammed earth. They examined both smooth and ribbed bars and demonstrated that ribbed bars exhibited up to 60% higher pull-out resistance. Furthermore, smooth bars, unlike ribbed bars, exhibited a sudden loss of load-bearing capacity without any post-failure phase. Bayesteh et al. [18] presented pull-out tests of various round bars. They examined smooth and ribbed steel bars, as well as ribbed GFRP bars, in two diameter variants, considering two different cement contents. They demonstrated that bond strength increased with cement content. The effect of increasing diameter in study [18] proved ambiguous: for smooth bars, bond strength increased with diameter, whereas for ribbed bars, it decreased. Zhou et al. [19] also investigated the effects of anchorage length, diameter, and compressive strength of stabilized soil on the bond of ribbed bars. Pull-out tests were conducted in situ for four bar diameters, using cement and lime as stabilizers. An increase in bond strength was observed with increasing mixture compressive strength. Increasing diameter resulted in greater strength and additionally influenced the failure mode (slip or soil-cement matrix fracture). Chen et al. [20] investigated the effects of cement content, water content, curing time, and ribbed bar diameter on bond strength. They determined the peak and residual bond strength values. They proposed a trilinear model for the bond stress–slip relationship, defining the elastic stress phase, the softening phase, and the residual stress zone. Zhang et al. [21] studied the influence of various design factors on bond strength and behavior. Strain gauges were placed on ribbed bars, and stress–slip curves were developed. Two failure modes were identified: splitting (sample fracture) and bar slippage. They noted that applying methods used for calculating anchorage in concrete significantly underestimated bond strength. Tests conducted on round or ribbed bars have limited utility in determining the contact conditions between steel sections and soil-cement mixes. This paper attempts to determine the bonding forces between a rigid hot-rolled flat bar and soil-cement, which more closely reflects the behavior of steel reinforcement profiles. The study considered the influence of initial roughness and oxide scale.

2. Materials and Methods

2.1. Soil-Cement Mixture

Soil-cement mix columns or panels are typically constructed in cohesive, low load-bearing soils. Therefore, clay, silt, and sand were included in the mix. The clay and silt were collected using a hand auger (Figure 1) from the Kraków–Częstochowa Upland. This region is characterized by diverse geology [22]. The sand fraction consisted of washed sand, commercially available in bags.

The collected clay and silt were then dried at 105 °C to a constant weight. The natural moisture content was also measured (Figure 2). To accelerate mixing and homogenize the mixture, the soil was crushed in a Los Angeles mill and then sieved.

The natural soil moisture content was 75.7% for clay, 85.5% for silt, and 15% for sand. The soil mixture of clay, silt, and sand was prepared in proportions of 50%, 30%, and 20%, respectively. The grain-size distribution curve of the soil used to prepare the mixture is shown in Figure 3. A cement content of 180 kg per cubic meter of soil-cement mix was adopted. The cement paste was prepared with a water-to-cement ratio of 1.0. Additional water was added to account for the natural moisture of the soil. The mass of each component of the soil-cement mix per cubic meter is presented in

Table 1. Composition of soil-cement mix.

	Density (kg/m3)	Natural Moisture (%)	Part of Soil-Mix (%)	Mass per Cubic Meter of Soil Mix (kg)
Clay	2100	75.7	50.0	731.0
Silt	1900	85.5	30.0	418.0
Sand	1750	15.0	20.0	343.0
Total Soil				1492.0
Cement				180.0
Water				180.0

.

2.2. Steel Bars

The tests were conducted on flat bars 30 mm wide and 4 mm thick, made of S235JR steel in accordance with the relevant standard [23]. To investigate the effect of surface condition on bond strength, three types of samples were prepared (Figure 4a): untreated (“R” series), brushed (“B” series), and ground (“P” series). The “R” series samples had the initial roughness of hot-rolled steel, along with a thin layer of oxidized scale on the surface. The “B” series samples were brushed to remove the scale layer; this treatment did not alter the surface roughness. The “P” series samples were ground with P220, P360, and P500 grit sandpaper; this treatment removed the scale and significantly reduced the surface roughness, resulting in a smooth surface finish. The roughness of each sample was measured using a Mitutoyo Surftest SJ-210 (Mitutoyo, Sakado, Japan) surface roughness tester (Figure 4b).

Table 2. Roughness measurement results of the R series.

Sample	Ra (µm)	Rq (µm)	Rz (µm)
R01	2.44	3.09	14.98
R02	2.03	2.58	12.30
R03	2.31	2.77	11.95
R04	3.25	3.96	17.43
R05	2.09	2.64	12.04
R06	5.08	6.19	28.53
R07	2.26	2.76	13.56
R08	2.95	3.69	17.71
R09	16.68	19.91	76.87
R10	19.35	22.02	78.00
Mean (µm)	5.84	6.96	28.34
Std. dev. (µm)	5.89	6.76	23.82
CoV (%)	100.74	97.12	84.07

,

Table 3. Roughness measurement results of the B series.

Sample	Ra (µm)	Rq (µm)	Rz (µm)
B01	2.87	3.53	17.03
B02	2.67	3.24	15.49
B03	2.30	3.03	17.34
B04	4.21	5.37	26.94
B05	2.54	3.21	16.63
B06	2.34	3.02	15.98
B07	3.24	4.14	21.87
B08	1.85	2.34	12.84
B09	2.48	3.02	14.80
B10	2.93	3.55	15.81
Mean (µm)	2.74	3.44	17.47
Std. dev. (µm)	0.58	0.74	3.66
CoV (%)	21.16	21.52	20.95

and

Table 4. Roughness measurement results of the P series.

Sample	Ra (µm)	Rq (µm)	Rz (µm)
P01	0.29	0.39	2.70
P02	0.43	0.62	3.83
P03	0.52	0.72	3.94
P04	0.21	0.31	2.58
P05	0.61	0.94	5.39
P06	0.62	0.82	5.37
P07	0.85	1.17	5.62
P08	0.72	0.94	5.18
P09	0.76	1.10	5.62
P10	0.41	0.54	3.68
Mean (µm)	0.54	0.75	4.39
Std. dev. (µm)	0.19	0.26	1.08
CoV (%)	34.60	35.07	24.51

present the roughness parameters Ra, Rq, and Rz, measured in accordance with the standard [24] for each sample. The roughness of each sample was measured at six points: three on each side along its axis. The results were then averaged. The coefficients of variation for the measurements of a single specimen (7–10%) were significantly lower than those for the entire series.

The measurements indicate significantly lower roughness for the “P” series samples. Within the “R” series, some samples (R06, R09, and R10) exhibited significantly higher roughness. It is likely that the bar from which they were cut originated from a different material batch. However, as they came from the same manufacturer and were purchased in the same store, and since macroscopic assessment does not provide strong grounds for exclusion, these samples were retained (Figure 4a). The remaining “R” series samples exhibited roughness values remarkably similar to those of the “B” series. The mean roughness values Ra, Rq, and Rz were 2.47 μm, 3.07 μm, and 14.28 μm, respectively, with coefficients of variation of approximately 15%.

2.3. Sample Preparation

Cubic samples measuring 150 × 150 × 150 mm were prepared with a centrally embedded flat bar. The samples were cured for 28 days under air-dry conditions. Due to the closed formwork structure, evaporation was limited, approximately corresponding to in situ curing of the soil-cement mix. Ten samples were prepared for each bar surface condition, and three samples without a bar were prepared to determine compressive strength. Since all samples were made from a single mixture, each sample name corresponds to the designation of the flat bar placed in it. Samples without a flat bar are marked with the designation “RC.”

2.4. Laboratory Tests

Three “RC” series samples were tested in compression using a Walter Bai (W+B) testing machine at a loading rate of 0.08 kN/s, corresponding to a compressive stress increase of 3.56 kPa/s. The bond strength was determined using a pull-out test. For simplicity, the force was applied vertically downward, and the bar was allowed to move freely downward through a cut-out in the base. The samples were tested in a W+B testing machine at a loading rate of 0.01 kN/s, corresponding to a bond stress increase of 0.98 kPa/s. The test setups for the compression and pull-out tests are shown in Figure 5.

3. Results

3.1. Compression Strength

All samples exhibited a failure mode characteristic of the uniaxial compression test of brittle materials. The test results—peak force P_max and compressive strength R_c—are presented in

Table 5. Compression test results.

Sample	Pmax (kN)	Rc (MPa)
RC1	34.27	1.52
RC2	29.25	1.30
RC3	30.98	1.38

.

The mean compressive strength was 1.40 MPa, with a coefficient of variation (CoV) of 6.61%. For 13 samples subjected to pull-out tests, the rod was removed without damaging the sample. These samples were then subjected to compression in the same manner as the “RC” series samples. The mean compressive strength was 1.27 MPa, with a CoV of 12.94%. The compression test results for both the “RC” series and the post-pull-out samples do not indicate population heterogeneity in terms of compressive strength, and the variance is typical for soil-cement mix [2].

3.2. Bond Strength and Stiffness

The failure mode for all samples was slip at the steel–soil-cement mix interface. The force–displacement characteristics were linear up to the peak value, and after exceeding the bond strength, they exhibited residual bond strength. To evaluate the failure surfaces, some of the specimens were fractured after the pull-out test. A typical debonding surface of the soil-cement, as well as the force–displacement curves, are shown in Figure 6.

Based on the force–displacement characteristics, the peak bond stress value τ_b (Equation (1)) and the tangential bond stiffness k_b (Equation (2)) were calculated within the range of 20% to 80% of the maximum stress.

$${\tau }_{\mathrm{b}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}{{\mathrm{A}}_{\mathrm{s}}}},$$

(1)

$${\mathrm{k}}_{\mathrm{b}}={\displaystyle \frac{0.8{\tau }_{\mathrm{b}}-0.2{\tau }_{\mathrm{b}}}{\mathrm{u}\left(0.8{\tau }_{\mathrm{b}}\right)-\mathrm{u}\left(0.2{\tau }_{\mathrm{b}}\right)}},$$

(2)

where P_max, A_s, u(0.8τ_b) and u(0.2τ_b) are, respectively, the maximum pulling force, the anchorage area of the bar (equal to 102 cm²), and the bar displacement values corresponding to 80% and 20% of the maximum stress. To ensure sample homogeneity, specimens with varying roughness were excluded from the “R” series (R6, R9, and R10). Descriptive statistics of the τ_b and k_b values are summarized in

Table 6. Descriptive statistics of τ_b and k_b.

Series	τb (kPa)				kb (kPa/mm)
	Mean	Std. Dev.	Min.	Max.	Mean	Std. Dev.	Min.	Max.
R	108.42	22.70	75.61	128.19	321.09	126.98	115.44	534.98
B	114.27	46.11	48.655	176.78	354.96	170.34	71.63	585.69
P	99.45	31.90	52.83	160.99	240.26	174.87	8.59	533.46

.

To assess the normality of the distributions of the variables τ_b and k_b, the Kolmogorov–Smirnov (K–S) and Shapiro–Wilk (S–W) tests were performed at the 5% significance level. The K–S test for both τ_b and k_b indicated that the null hypothesis of normal distribution should be rejected for all series. However, the Shapiro–Wilk test suggested that the null hypothesis of normality could not be rejected for all variables in the series, except for τb in the “R” series. Figure 7 and Figure 8 present the Q-Q plots for τ_b and k_b.

The results of the normality tests were not conclusive. Therefore, to compare differences between the series, a nonparametric test that does not require the assumption of normality was applied. The Mann–Whitney U test (also known as the Wilcoxon rank-sum test), which is the nonparametric counterpart of the two-sample t-test, was selected [25]. The test probabilities for the null hypothesis that the values originate from the same distribution are presented in

Table 7. Wilcoxon rank-sum test probabilities for τ_b.

Series	R	B	P
R	-	0.8125	0.5362
B	0.8125	-	0.4727
P	0.5362	0.4727	-

and

Table 8. Wilcoxon rank-sum test probabilities for k_b.

Series	R	B	P
R	-	0.8125	0.2295
B	0.8125	-	0.1620
P	0.2295	0.1620	-

.

3.3. Correlation Analysis

A correlation analysis was performed to verify the relationship between mechanical properties and roughness. First, a correlation analysis was carried out between the roughness parameters. The correlation matrix is presented in Figure 9.

Excluding outliers (R6, R9, and R10) did not significantly affect the correlation strength (Figure 10).

The Ra, Rz, and Rq roughness parameters are strongly correlated, indicating that they describe very similar aspects of surface topography and the specimens do not contain any large, individual defects (deep scratches, pores, burrs). These parameters convey similar information and can be treated as mutually interchangeable in assessing surface quality [26,27]. The correlation analysis showed that Ra, Rq, and Rz for the tested samples were linearly dependent; therefore, the roughness–strength and roughness–stiffness correlations were investigated only for the Ra parameter. This parameter is the most suitable due to the surface topography of the specimens as well as the potential application. In industrial applications, soil-cement columns and walls are reinforced with I-beam sections of considerable length (on the order of 10 m) and very large surface areas; therefore, an averaged roughness measure, the least sensitive to local irregularities, is of particular importance. The correlation matrices are presented in Figure 11 and Figure 12.

4. Discussion

Compressive strength tests on cubic specimens demonstrated that the soil-cement mix was homogeneous in terms of strength characteristics. The obtained characteristic compressive strength of 1 MPa is typical for low-strength soil-cement mixtures used in structures requiring a low degree of reinforcement [1]. Compression tests conducted on samples after the pull-out test showed an average strength 9% lower, which can be explained by weakening of the hole after flat-bar removal.

The bond behavior of reinforcing bars in concrete is determined by the chemical attraction of particles at the steel–concrete interface (adhesion), by mechanical interlocking, and by friction [11]. As the surface roughness increases, the relative contribution of mechanical interlocking and friction at the steel–soil-cement matrix interface becomes more pronounced. Regardless of the method of preparing the flat-bar surface, the failure mode was slip, with no fracture in the soil-cement mix. Smooth rods in the referenced studies [17,18] behaved in a similar way. The failure mode corresponds to the pull-out tests on smooth round bars [17,18], but quantitative comparisons of the bond-strength values are unjustified due to differences in bar geometry and soil-cement mix properties. Furthermore, 90% of the flat bars, regardless of surface preparation, exhibited residual bonding [20], which is related to friction on large contact surfaces (significantly larger than in the case of round bars).

Roughness tests revealed sample heterogeneity in series R, which was addressed by excluding samples R6, R9, and R10 from the comparative analysis of series R, B, and P. The results of the normality tests for the strength and stiffness variables were ambiguous: the K–S test rejected the null hypothesis of normality for all series, while the S–W test rejected it only for the strength of series R. Due to this ambiguity, parametric tests were abandoned in favor of the nonparametric Wilcoxon rank-sum test. The high test probabilities do not provide a basis for determining significant differences between the series in terms of either strength or stiffness. The tests conducted did not show any significant influence of scale on the bond parameters. Comparison of the P series with the B series does not indicate any significant effect of initial roughness on the strength. Noteworthy are the results for samples R9 and R10, which had outlying Ra roughness values of 16.68 μm and 19.35 μm, respectively. The bond strength of sample R9 was 215.67 kPa and that of sample R10 was 256.60 kPa, both much higher than the average bond strength for the R series of 108.42 kPa. This indicates that roughness in the 0–5 μm Ra range has no significant effect and that the effect increases only above this value. For low Ra roughness values (0–5 μm), adhesion determines bond strength, as in the case of smooth bars [17,18]. With increasing roughness, the influence of mechanical interlocking increases, which explains the high bond strength of samples R9 and R10.

The value of the Pearson correlation coefficient (r) between bond stiffness and Ra roughness indicates no correlation. The relationship between roughness and bond strength appears significant, but analysis of the scatterplot indicates that the correlation strength is determined by samples R9 and R10. Excluding these samples caused the r coefficient to drop to 0.33. The correlation analysis confirms that for small roughness values (Ra below 5 μm), no effect on bond strength was observed.

5. Conclusions

The tests conducted showed no significant effect of scale on the strength and stiffness values. Samples with Ra roughness below 5 μm exhibited similar strength and stiffness values, which can be attributed to the dominant effect of adhesion. The determined strength values of 100 kPa can be considered a conservative estimate of bond strength limits, as they were obtained from samples with low roughness in a soil-cement mix with low cement content. As Ra roughness increased to 15–20 μm, a two-fold increase in bond strength was observed. When samples with outlying roughness were included in the correlation analysis, a strong relationship between roughness and strength was observed.

The experimentally determined bond-strength values were characterized by high variance, which makes drawing firm conclusions difficult. However, the results presented in this article provide a basis for a power analysis of the test. Further research on the bond phenomenon in soil-cement mixes should be conducted on larger sample sizes. Developing a reliable model of the contact between steel profiles and soil-cement requires further research that considers different cement contents and soil fractions. As large-sample studies are proposed, future research will require careful experimental design [28,29]. The promising correlation between strength and higher roughness values suggests conducting tests on samples with artificially increased roughness [30] or on surface-corroded bars. The study requires testing for roughness values of Ra 10 μm and higher, up to Ra 1000 μm—a value corresponding to the grain diameter of the sand fraction. This will make it possible to investigate the effect of mechanical interlocking and friction on bond strength. Tests on mixtures with varying cement content will enable the assessment of the influence of adhesion. Moreover, observation of interfacial phenomena will additionally require the development of a pull-out test procedure allowing the use of computed tomography [31] to evaluate the failure surface and to identify the mechanism determining bond strength. Investigating the effect of stresses normal to the shear surface on bond strength and stiffness will be crucial for developing a two-parameter contact model that can be used in numerical simulations.