Abstract
In this study, a mixing procedure of sulfur-rich soil and cement-based binder to enhance the soil’s unconfined compressive strength (UCS) was tested in field conditions for geotechnical applications. The focus was to evaluate uniformity of industrial size soil-binder mixture, blended by existing method. This paper outlined sampling strategy and the number of samples needed for a valid uniformity evaluation. Moreover, this study emphasized the difference between field mixing and laboratory mixture preparation by comparing UCS of stabilized soil samples in the field and UCS of corresponding samples mixed and prepared in the laboratory environment. In the field, soil and cement were blended in two to four stages with 5% and 7% cement—the percentages being based on the soil’s dry weight under field conditions. Samples were taken from the field mixtures after each stage. Since the number of samples needed to be representative of mixture characteristics for large-scale mixing is not standardized, this field experiment included two phases. The first phase was dedicated to finding a sampling strategy for a large soil pile along with measuring UCS of collected samples. In the second phase, sample collection was conducted based on the results of sampling strategy from the first phase. In the laboratory, samples with percentages of binder similar to the amount of binder in the field were also prepared. Both field and laboratory samples were prepared using the tapping method in the laboratory for UCS test. Samples were cured under similar conditions for 28 days. The results showed that the uniformity of mixture improved after each additional mixing stage. In addition, while spots with low UCS were observed in the second mixing step, these spots were eliminated in the third mixing step, and results of the UCS tests were comparatively uniform. Moreover, comparison of the samples revealed that the UCS of the laboratory mixture is higher than that of the field mixture. This showed that even though the UCS is a good standard for comparing the strength of different soils stabilized with different percentages or types of binders in the field mixing, the actual strength of the stabilized mixtures under field circumstances is lower than that in laboratory prepared mixtures.
1. Introduction
Sulfur-rich soil, commonly known as sulfide soil, is clayey silt soil that is widely found in the coastline areas of Sweden and Finland. Sulfur-rich soils containing metal sulfides but not yet oxidized or exposed to oxygen with the potential to generate an acid leachate are called potential acid sulfate soils and referred to as sulfide soil. Acid sulfate soil is one that has already been oxidized and is already acidic. Among the sulfide minerals found in this soil, iron disulfide marcasite known as pyrite (FeS2) is the most common. The examples of consequence ofacid generation in the oxidized soil are the mobilization of metals which affects aquatic ecosystems and degradation and corrosion of steel infrastructures. Furthermore, cement mixtures, including concrete, cannot be held in an acidic environment with a pH value below 3. Due to high deformability and lack of stability, it is often excavated and replaced by geotechnically approved materials such as gravel or well-graded soils. Sulfide soils that have been excavated are potential acid sulfate soils that have not yet been oxidized. Furthermore, since these soils have acid-producing ability as a consequence of their chemical composition, excavated soils are transported to landfills with particular conditions making the management of sulfur-rich soil laborious and costly. One solution for reducing the volume of excavated soils requiring landfilling is reusing soils in geotechnical applications as construction material. Since the mechanical properties of these types of soil are not adequate for construction, soil stabilization is necessary before applying them. Soil stabilization is one method to enhance the mechanical properties of soil with low workability and strength. Adding binders to the soil is frequently used to modify soil’s engineering properties. The cement-soil mixing technique is widely employed in field construction for enhancing strength and improving compressibility of soil such as silty clay, clay, and sewage sludge [1–5]. Unconfined compressive strength (UCS) is one of the most important soil parameters that describe soil strength, and it can be improved by adding cement to the soil. Al-Jabban et al. [6] proved that the UCS increases considerably by adding a small amount of cement to the soil sample. Multicem—a mixture of cement kiln dust (a by-product of the cement manufacturing industry) and cement—is another alternative to cement for improving the physical characteristics of the soil. Rothhämel et al.’s [7] research showed that adding Multicem to the silty sand improved the properties of the soil after freeze and thaw cycles.
While several studies show the positive effect of binder on the UCS of silty soil in a controlled laboratory environment [8–10], only a few studies investigate the effects of soil and binder mixing procedure on the UCS improvement. Deep mixing is one of the methods used widely for in situ stabilization of soil layers [11, 12], but none of these have clearly studied excavated soil mixing procedure. Although adding a binder to the soil effectively modifies soil characteristics, there is no standard technique for mixing excavated soil and binder to enable reuse of the soil.
Mixing soil and binder on a large scale requires a homogeneous mixture in order to ensure that the stabilization results are achieved. There is a belief that modern equipment has eliminated mixing problems. It is true that cement can be mixed with most soils, including clays, under ideal conditions [13]. A question arises concerning how well cement and soil are mixed. A number of tests have shown that the field strengths of soil cement rarely match laboratory strengths [14]. Insufficient mixing is believed to be the cause of this problem. As a measure of field mixing efficiency, the British used the ratio of field strength to laboratory strength. There has been a variation in mixing efficiency depending on the mixer type and soil type, but a typical value is 60%. It is noted that if the mixing efficiency could be increased to 80%, 30% less cement would be required for the same strength [13]. Also, assessing the homogeneity of a soil pile is challenging. Larsson [12] studied the distribution of binder in the mixing columns in deep mixing project. The study showed that by applying statistical analysis and calculating mixing indices, the mixing quality is measurable. The drawback of the method is that a large number of samples are needed for the evaluation of mixing quality.
Since one main effect of mixing binder with soil is to improve strength of the soil, the quality of the mixture can be evaluated by the strength of the collected samples from the mixture pile. The stabilized mixture homogeneity can, therefore, be assessed by measuring the strength of soil pile samples, i.e., characteristics of collected samples describe the soil pile properties. Therefore, it is necessary to perform soil sampling that ensures that the samples are collected evenly across the soil. Another important point that needs to be considered in sampling is the minimum number of samples. The minimum number depends on scale of the pile and the heterogeneity of mixtures. The number of samples is as essential as the sampling technique for a sampling procedure to be representative of the mixture.
The overall aim of the study was to evaluate homogeneity of soil and binder mixture mixed with full-scale equipment, measuring UCS of mixture as a quality criterion for the mixing operation. These specific objectives were addressed:(i)Choose and evaluate a sampling strategy to assess the homogeneity of the soil pile.(ii)Determine the effect of the binder dosage on the UCS and number of mixing steps.(iii)Assess the difference between the UCS of laboratory samples and the UCS that can be achieved in the field.
2. Materials
2.1. Method of Soil Characterization
The fieldwork was conducted at Dåva Landfill and Waste Center in Umeå AB. The soil was from the construction site of a railway terminal close to the landfill. The soil excavated from the site was brought to the waste disposal site.
The binder chosen for this experiment was Multicem, which was ordered from Cementa. Multicem contains 50% cement and 50% cement kiln dust (CKD) and is a by-product of cement manufacturing [15]. In the production of Multicem, a lower amount of raw material has been used. This reduces CO2 emissions for the finished product compared to conventional cement.
The soil’s chemical composition was measured according to SS EN ISO 17294 and SS EN ISO 11885. The loss on ignition was measured on samples taken from the site following ASTM standard (D-13, 2013). The moisture content of the soil was measured based on the ratio of water to the solid mass of the soil (ASTM D4442-16). The consistency limits of the soil, including plastic and liquid limits, were measured according to [16] and the fall cone method [17]. The particle size distribution of the soil was obtained from both sieve analysis and sedimentation following the Swedish standard [18].
2.2. Soil Characterization
This soil was classified as sulfur-rich soil with a loss on ignition of 3% at 450°C, and the chemical composition is presented in Table 1. The original water content was 50%. Plastic and liquid limits were measured to be 23% and 33%, respectively, with PI = 10%, which categorized the soil under medium plasticity type of soils [19]. The particle distribution curve is presented in Figure 1.
3. Experimental Plan
In this field experiment, batches with two binder percentages were mixed by bucket mixer in steps. This fieldwork delved into batches’ homogeneity after each mixing step by collecting samples and comparing the UCS of collected samples within mixing steps. The field mixing was divided into two phases to develop a strategy to obtain representative samples to assess the quality of the mixture at the scale of this experiment. The first phase was used to determine the pattern of sampling and number of samples representative of industrial size soil pile along with testing UCS of samples to determine the effects of mixing steps on uniformity of mixture. The second phase included applying the most efficient sampling technique based on results of first phase and changing the amount of binder to study the effect of binder percentage on mixing procedure of industrial scale soil-binder mixture. The results of the mixing procedure were evaluated based on the results of the UCS tests for both phases. Eventually, UCS of samples from field was compared with equivalent laboratory experiment. This comparison provided the opportunity to picture the field results based on laboratory data and optimize mixing procedures.
3.1. First Phase: Determination of a Sampling Strategy
3.1.1. Preparing the Soil-Binder Mixture
Between 15 and 20 tons of soil were transferred to the site (Figure 2(a)). Initially, soil samples were homogenized using ALLU bucket mixer with volumes 2.3 or 2.7 m3 (Figure 2(b)). Pure soil samples were taken from the batch for further laboratory experiment and soil characterization, including preparing laboratory mixture, as well as determining particle size distribution, consistency limits, and chemical composition. A three-step mixing process was used to blend soil and binder. An amount of Multicem equal to 7% on dry weight basis (180 kg) was added as a binder. Multicem was spread on top of the weighted soil pile prior to mixing procedure (Figure 3(a)). As shown in Figure 3, a mixing step consists of mixing by mixer and sampling which was repeated three times for the first phase. To conduct the mixing experiment, 5.1 tons of the batch was weighed from the soil pile shown in Figure 2(a). Sequentially, samples were collected from the soil pile after second and third mixing steps.
3.1.2. Soil Sampling and UCS Sample Preparation
Figure 3(c) shows how the soil pile was divided after the second and third mixing steps. This sectioning strategy was selected for samples to be evenly distributed across the soil pile. The mixture pile was divided into 5 sections, and from each section, mixture sample was taken and transferred to the laboratory in buckets to prepare UCS samples. The number of UCS samples prepared from this phase is presented in Table 2. The soil-binder mixture was compacted in the tubes in 5 layers (Figures 4(a) and 4(b)). Since water content of mixture was extremely high, it was impossible to compact the samples using a light proctor hammer. As an alternative method, each layer was tapped 30 times by hand on the working surface to ensure that the soil was compacted and to reduce voids. This molding technique has been studied by the Tokyo Institute of Technology (TIT), the Sapienza University of Rome (UR), the University of Coimbra (UC), and the Swedish Geotechnical Institute (SGI) to investigate the influence of molding technique on density and unconfined compressive strength of soil [20]. Samples were cured for 28 and 90 days, and the samples were tested in the uniaxial compression test apparatus with the deformation rate of 1 mm per minute (1% of height) after each curing time [21] (Figure 4(c)).
(a)
(b)
(c)
3.1.3. Laboratory Experiment
A soil sample from the field was brought to the laboratory in order to investigate the stabilization results in small scale. The laboratory samples were prepared by mixing soil with 6% and 8% Multicem in laboratory environment. A stand mixer was used to blend soil and binder after homogenizing the pure soil in small batch. The samples were prepared in the same tubes as the field experiment by the tapping method. Samples were cured in the laboratory together with field samples.
3.1.4. Stratified Sampling Method for Industrial Size Mixture Experiment
To decide the number of representative samples for this soil pile, the one-way analysis of variance (ANOVA) was used to examine the possible relationship between the number of samples and other parameters, including mixing time and number of sections. The one-way ANOVA was used to determine whether there were any statistically significant differences between the means of three or more independent (unrelated) groups. For this purpose, the differences between the UCS values of each section were examined after second and third mixing times. The one-way ANOVA was used for five groups of data—in this case, five sections—to obtain information regarding the relationship between the number of samples taken from each section and to determine whether there was a significant difference between the UCS results of each section. Microsoft Excel was used to calculate the value, which is a statistical measurement used to validate a hypothesis against observed data. The field experiment was continued after analyzing the results of first phase which led to changing the number of collected samples for the second phase.
3.2. Second Phase: Assessment of the Binder Dosage
The primary characteristics of soil are also applied to the second phase. The moisture content was measured for the second phase, and no significant differences were found. Two main differences between the first and second phases were the amount of binder added to the mixture and the number of sampling sections. Table 3 shows the difference between the first and second phases. The binder amount reduced to 5% and sampling section increased to 12 (Figure 5) according to the conclusion reached in the first phase. The number of mixing steps also increased to 4 times in order to investigate the effect of further mixing on the homogeneity and strength of stabilized soil pile. Single UCS samples were prepared from each section after mixing steps in order to analyze homogeneity of mixture pile by comparing UCS of these samples. The samples were prepared according to the same experimental procedure in the first phase, and they were cured for 28 days.
4. Results
4.1. Results of the First Phase
Figure 6 shows the UCS results of the second and third mixing steps’ samples cured for 28 days. The five different sections of the pile are illustrated by specific colors and numbers. For instance, 3-2 means sample number 2 from section number 3. Sections in second step of mixing are numbers 1 to 5, and sections for third mixing step are numbers 6 to 12. While the second mixing step results indicate that there is one spot with lower UCS in the soil pile (2-1 and 2-2, in orange), the differences between the results of different sections after the third step of mixing are comparable. Figure 7 shows the UCS results of second mixing step samples after 90 days of curing. This figure indicates that increasing the curing time improved the UCS of all samples to some extent. Samples 2-3 and 2-4—prepared from the same spot with low UCS in Figure 6—showed lower strength growth in comparison with other samples. This can prove that less binder was mixed with the soil in this spot. The average UCS test results for the second and third steps of mixing after 28 days of curing are 46 kN and 43 kN, respectively. Moreover, Figure 6 shows that even though there are differences between the results of UCS between different sections, the differences between samples from the same section are negligible. To verify this statement, one-way ANOVA was utilized. Since there were a larger number of samples taken from the third step of mixing, the results of the UCS test of these samples were used for statistical analysis to determine the best way to sample the industrial size soil mixture pile. The calculation of the value for the third mixing step showed a significant difference between the average UCS results of the sections ( value ≤ 0.05). The same conclusion was obtained evaluating second mixing step that UCS of samples within the sampling section is approximately equivalent. By performing the one-way ANOVA method within triplicate samples for each section for the third mixing step, the result of value is greater than .
Because the hand tapping method was chosen for sample preparation, errors likely were introduced to the sample compaction. To demonstrate that samples were uniformly compacted, the density of each sample after each curing time was measured. As the black lines in Figures 6 and 7 show, the variation of the sample’s density is less than 0.05 g/cm3.
4.2. Results of the Second Phase
Based on assessment of sampling strategy in the first phase, the second phase of field experiment was planned, and the number of sampling section was increased to 12 sections. Also, it is concluded that single UCS sample will be representative of strength characteristic of each section. Figure 8 shows the UCS of samples prepared in the second phase of field experiment. Similar to the first phase, a low UCS spot was observed (sample no. 1) in the second mixing step. Histogram of probability distribution of UCS results was plotted for each mixing step using mean and standard deviation to assess the effect of mixing steps on the homogeneity of samples after each mixing step (Figure 9). As shown in this figure, the averages of UCS for the second and third steps of mixing are close, and this value is lower in the fourth step of mixing.
4.3. Comparison of Laboratory and Field Results
Comparison of the average UCS of field samples after each mixing steps and the laboratory results is shown in Figure 10. The UCS obtained with laboratory mixing is higher than that of field mixing. Since the amounts of added binder to the mixtures are different from the field, the UCS for 5% and 7% binder is interpolated based on trendline of laboratory results. The trendline is assumed to be linear since the difference in the binder percentages is low in laboratory mixtures. Initially, the binder percentage was chosen to be the same as the laboratory experiment, but a recalculation based on measured weights in the field resulted in 7% and 5% binder in the batch. This difference between the binder percentages in the field experiment and the laboratory mixtures happened due to limitations in weight measurement facilities in the field. There was a time limit that prevented from repeating the laboratory test based on correct field binder dosages. The predicted values of the laboratory UCS are calculated to be 108 kPa and 174 kPa for 5% and 7%, respectively, which are 80% and 75% higher than field experiment UCS averages.
5. Discussion
The main goal of mixing binder and soil is to enhance the strength of soil. Sulfur-rich soil originally has low bearing capacity due to high water content and poorly graded particle distribution. The UCS test shows that adding Multicem improves soil strength. The hydration reaction of binder leads to the formation of different components in soil including calcium silicate hydrates (CSH), calcium aluminate hydrates (CAH), and calcium aluminum silicate hydrates (CASH) which increase the binding between soil particles. The strength develops with the increase of curing time as intercluster bonding reactions between cement-based binder and soil are time-dependent. The UCS value was selected as a parameter that indicates how well the soil and Multicem are mixed and if the mixture is homogeneous. The initial assumption was that if the soil and cement are properly mixed, the UCS of mixture will improve and UCS values of samples of each section will have close values.
Comparing density of samples prepared with the tapping method showed that samples are equally compacted and the differences between UCS values of samples are not related to compaction method and density of samples. Therefore, although the tapping method is not a standard sample preparation method, this method proved to be a convenient and efficient method for working with high water content soils for laboratory sample preparation.
For solid material with sampling size of 30 tons or less, taking five to seven samples are suggested [22]. In the first phase, the soil pile was divided to 5 sections according to NordtestMethod [22]. Statistical analysis of sampling results in the first phase has shown that there was no significant difference between the UCS values of samples within a section but there were differences between the UCS results of the five different sections. Based on the one-way ANOVA analysis of results in the first phase, it was concluded that samples must be taken from different sections to evaluate the homogeneity of soil pile regarding the strength improvement. The value comparison showed that there was no significant difference between the UCS test results within a section. According to the conclusion reached in the first phase and ANOVA Analysis, each section can represent the characteristics of the specific location of the mixture pile. Therefore, the number of sampling sections increased to 12 and single samples were prepared from each section in the second phase.
The field experiment was conducted in two different phases with different Multicem percentages for each phase. The pure soil behaved like liquid before mixing with binder. It can be explained by its water content being higher than the liquid limit of soil. Adding Multicem increased the workability of the soil in two ways. The first is the immediate effect of Multicem by absorbing water and changing the consistency of soil. The second is the hydration reaction of binder which consumes water and produces heat which leads to moisture reduction in the final mixture. The latter has a minor effect on batch consistency in comparison with the first. No differences in the batch consistency caused by adding a higher percentage of Multicem to the batch were observed. Also, other environmental conditions may affect the measured consistency, e.g., the fieldwork was done at two different temperatures. While the second phase of the field experiment was carried out at −15°C, the temperature was above zero during the first phase. The consistency of the second batch could be affected by the frozen water in the soil. Microsoft Excel was used to conduct a simple linear regression to compare the results of the UCS test after each mixing step and to determine how much mixing was required for this particular batch size. This analysis considered UCS test results to be dependent on mixing steps. In order to spot if the dosage of Multicem creates difference between phases, this analysis was repeated for the data of both first and second phases. Mixing step was considered as an independent variable, and the UCS of the sample was considered a dependent variable. The UCS of the samples was then predicted as a function of mixing steps.
The homogeneity of soil-Multicem mixtures was assessed by comparing the UCS test results of samples within each mixing step (2nd and 3rd and 4th (second phase)) to determine the influence of increasing mixing steps. Figure 11(a) shows that predicting the UCS value only based on number of mixing steps results in high errors for 7% Multicem (first phase). This indicates that predicting UCS value based only on mixing steps results in very low accuracy (R2 = 0.02). However, the prediction model (the black line in the graph) shows decreasing UCS values as the mixing steps are increased.
(a)
(b)
The same analysis was repeated for the second phase of the field experiment. Figure 11(b) shows the predicted value of UCS for the batch with 5% Multicem. Since the R-squared value is low (R2 = 0.03), the prediction of UCS value based on mixing steps as a variable will be associated with a high risk of errors. While R2 is low in both batches with 7% and 5% Multicem, the prediction of results is comparable, which means both predictions show a minimal decrease of the predicted values after each mixing step.
Another point that needs to be considered corresponds to water retention characteristics of silty material exposed to intense mixing. As shown in Figures 11(a) and 11(b), the predicted UCS (red dots) is decreasing by increasing mixing steps. It means even though the homogeneity of samples improves by adding mixing steps, the strength of samples reduces. Mäkitalo [23] explains that intensive mixing of green liquor dregs and soil releases bound water and prevents mixture from compaction. Same phenomenon can cause decrease of average UCS by increasing mixing steps.
Figure 12 shows the difference between the measured residual strengths and the predicted UCS from the regression line for the data of both phases of field experiments. By comparing residual values of the second and third steps in Figure 12(a), it is indicated that the predicted value in the third step is closer to the measured values. Therefore, UCS values of batch are more uniform when there are more mixing steps. Similarly, in Figure 12(b), the residual values are reduced by increasing the mixing step for the batch with 5% binder in the second phase. According to the UCS results, the residual distribution in the 4th step of the mixing process is approximately even and the variance is approximately ±10. The range of results is highest in the second step of mixing in first phase (Figure 12(a)) showing that the UCS values of samples prepared from the second step of mixing are less uniform. The lowest variation is observed for the fourth mixing step. Previously, it was shown in Figure 9 that the variation of the results decreased when the number of mixing steps increased. It is shown in Figure 9 that in fourth step of mixing, the normal distribution graph has the closest shape to a symmetrical bell-shaped curve meaning that the measured UCS value for the fourth step of mixing is relatively equal to its average UCS value in comparison with other two mixing steps.
(a)
(b)
There are two main reasons to explain the notable differences between UCS values of samples from field and laboratory shown in Figure 10. First, as shown in Figure 3, the binder was dispersed on the top of soil pile and scattered around, and not all of the binder was mixed with the soil. It can be concluded that the percentage of mixture was less than what was measured at the first place. Second, the mixing procedure in the field was not as effective as in the laboratory because size of the sample in the laboratory was significantly smaller than the field batch. Thus, it is possible to achieve a well-mixed and uniform batch by using a stand mixer in laboratory. These results showed the importance of finding a correlation between different variables affecting samples’ physical characteristics including porosity and UCS to help predicting the strength behavior of stabilized soil in both field and laboratory.
6. Conclusion
The homogeneity of the mixture was evaluated by the UCS of the samples taken from the field mixture and used as a basis of the homogeneity assessment. The field experiment showed that increasing mixing steps will increase the homogeneity of samples to a certain extent.
Samples were mixed with 7% of binder, and the mixture pile was divided into 5 sections for sampling purpose. Samples taken from a specific section of the soil pile showed consistent UCS results, while results differed significantly when taken from different sections. This is because the soil at that specific section mixed with similar amount of binder. The amount of binder is relatively different from other sections within each mixing step as a result of mixing procedure efficiency. Therefore, it is suggested that increasing the number of sampling sections provides better representative samples and better interpretation of mixing efficiency. Results show that collecting samples from more sections and preparing single UCS samples provide representative results for this specific soil batch size.
Based on the results of this field experiment, the following conclusions can be drawn:(i)Reducing binder did not show measurable effect on uniformity of mixtures during mixing procedures. The UCS of the samples showed the same pattern for both mixtures.(ii)Increasing mixing step from two steps to 3 steps showed significant effect on improving the uniformity of UCS results.(iii)Results indicated that sections in the soil mixture pile with significantly low UCS developed while mixing the batch for the second time. These spots were not observed after the third mixing.(iv)An UCS average reduction was observed after each mixing step. This reduction was not significant compared to the increase in homogeneity of the mixture.(v)Comparison between UCS values of field and laboratory mixing showed that latter produced significantly higher strength than the former. Prediction of field results based on laboratory mixed samples should involve high safety factors.
To better predict the effect of binder addition on the strength of the soil in constructions, future research should address the factors affecting UCS of stabilized sulfur-rich soil in the laboratory environment and investigate the correlation between those factors and measured UCS. Enhancing knowledge concerning these controlling factors would permit a better prediction of the strength of stabilized soil mixtures in the field.
Data Availability
The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This project was financed by the European Regional Development Fund via the Interreg Botnia-Atlantica program for the project Sustainable treatment of coastal deposited sulfide soils (STASIS). Staff of Dåva D.A.C waste center are thanked for preparing field facilities and performing the field experiment.