Assessing the progress of internal instability of fill dam material in South Korea through long-term seepage tests

Abstract

Several studies investigating the internal instability of well-graded soils have suggested that well-graded soils show characteristics and developments different from those of gap-graded soils. Moreover, the cause and progression of long-term internal instability in well-graded soils have not yet been established. This study aimed to address this knowledge gap by conducting long-term seepage tests to analyze the progress and cause of the internal instability of well-graded soils generally utilized in fill dams in South Korea. The test results revealed that well-graded soils in internally stable conditions exhibited clogging due to fine particles, reducing overall permeability. A similar phenomenon was identified under internally unstable conditions. The subsequent unclogging process resulted in a rapid increase in permeability and erosion rate with the collapse of the soil structure. Consequently, the internal instabilities of well-graded soils progress through suffosion. Based on the test results, the mechanism of internal instability and its progression in well-graded soils were proposed.

Introduction

Internal erosion is a critical process affecting the stability of fill dams. Foster et al. (2000) reported that 57 of the 126 cases of dam failures abroad were caused by internal erosion, accounting for 45% of the total number of failures. Suffusion and suffosion, also known as internal instabilities, are forms of internal erosion. Suffusion involves the selective erosion of finer particles from the matrix of coarser particles of internally unstable soil. The finer particles are removed through voids between larger particles by seepage flow, leaving behind a soil skeleton formed by the coarser particles. Suffosion is a process that results in volume changes together with selective erosion (Fell et al. 2015; Reclamation and USACE 2019). The erosion of finer particles increases the permeability of the soil, resulting in a decrease in shear strength and changes in hydraulic conditions. This erosion may cause various types of dam failures, such as backward erosion piping, crest settlement, overtopping, and downstream sliding (Chang and Zhang 2013b; ICOLD 2017; Lee et al. 2021).

Generally, gap-graded soils with two distinct grain sizes are vulnerable to internal instability, which is evident through various manifestations of suffusion. These include the loss of a substantial quantity of fine particles and an increase in the coefficient of permeability (k) and void ratio (e). Several researchers have studied the causes and progress of internal instability in gap-graded soils (Skempton and Brogan 1994; Moffat and Fannin 2006; Bendahmane et al. 2008; Chang and Zhang 2013a; Luo et al. 2013; Ke and Takahashi 2014). In particular, Luo et al. (2013) performed both short-term and long-term suffusion tests on gap-graded sandy gravel, analyzing and comparing the characteristics of internal instability progression in both tests. However, most fill dam materials in South Korea are comprised of well-graded soil formed by the decomposition of granite and gneiss (NDMI 2013). Although such soil type is not only limited to the fill dam materials in Korea but also has been used worldwide for dams (Li 2008; Moffat et al. 2011; Rönnqvist and Viklander 2015), there have been only a few studies conducted on short-term seepage tests with well-graded soils, and long-term seepage tests have not been thoroughly investigated (Sterpi 2003; Wan 2006; Li 2008; Moffat et al. 2011; Israr and Indraratna 2019; Liu et al. 2021). The majority of these studies have primarily focused on post-occurrence phenomena rather than delving into the processes and causes of internal instability. Furthermore, in previous short-term test results (Lee et al. 2022), the internal instability of well-graded soil (the same specimen used in this study) showed sudden increases in k and soil discharge, only after a progressive reduction in k, revealing an internal erosion process different from that of the gap-graded soils. This phenomenon was also observed in experimental studies performed by Li (2008) and Liu et al. (2021).

The “short-term” suffusion tests are the experiments with a stepwise increase in hydraulic gradient until the soil reaches its failure, with each step having a defined duration (Li 2008; Luo et al. 2013; Liu et al. 2021; Lee et al. 2022). This method offers the advantage of determining the internal instability in a relatively short time frame and providing an approximate estimation of the hydraulic gradient at which the internal instability initiates. However, short-term test results may have limitations in assessing the progression of internal instability. The stepwise increase in hydraulic gradient over a short period can be influenced by the hydraulic gradient applied during the preceding stage, potentially affecting the internal erosion resistance of the soil. On the other hand, while “long-term” suffusion tests have the disadvantage of requiring a significant amount of time due to the need for conducting tests under each hydraulic gradient, they offer the advantage of measuring eroded soil volume, flow rate, and settlement under a constant hydraulic gradient until internal instability occurs. Consequently, long-term tests are frequently employed to investigate the evolution mechanism of suffusion (Luo et al. 2013). Particularly, since internal instability in well-graded soils tends to manifest slowly and progressively over time (Fell et al. 2003), long-term tests are more suitable for understanding the progression of internal instability and can aid in predicting the lifelong stability of actual fill dams.

This study conducted long-term suffusion tests on well-graded soil with a representative particle size distribution of the fill dam materials in South Korea. The “long-term” is used relatively, in contrast to the “short-term” tests, and it involves maintaining a constant hydraulic gradient until internal instability occurs, or for a maximum of 576 h. Throughout these tests, the amount of eroded soil, flow rate, and settlement are measured over time under various testing conditions of relative densities and hydraulic gradients. Based on the test results, the influence of the hydraulic gradient and relative density on the internal instability of well-graded soils and the cause for the variability in k and settlement were analyzed. In addition, the mechanisms underlying the progression of internal instability in well-graded soils were investigated.

The main objectives of this study were as follows: (1) to analyze the influence and causes of hydraulic gradients and relative density on the internal instability of well-graded soils, (2) to investigate changes in the permeability coefficient and settlement, and (2) to determine the mechanism underlying the progression of internal instability in well-graded soil.

Testing materials

Well-graded silty sand (WG soil), a material commonly available in South Korea, was selected as the soil sample for the seepage tests. WG soil is a natural silty sand decomposed from granite and has a particle size distribution representing the fill dam materials in South Korea. Figure 1 shows that the particle size distribution of the WG soil is within the range obtained from nine different domestic fill dams (36 particle size distributions from nine fill dams in operation, as investigated by the Korea Rural Community Corporation). Well-graded soils in general have a linear particle size distribution with uniformity coefficient (${C}_{u}$) values larger than 4, whereas the well-graded soils used as fill dam materials in South Korea have a wide range of particle sizes, with ${C}_{u}$ values larger than 20. WG soil was classified as clayey, silty sand (SC-SM) according to the Unified Soil Classification System (USCS), with fine content (sieve no. 200) of 40% after eliminating particles with diameters larger than sieve no. 4. The clay content of the sample is approximately 4% exhibiting low plasticity index of 5.9%. In Table 1, soil classification by USCS, ${C}_{u}$, and specific gravity of WG soil are similar to those of fill materials in Korea. However, the clay content and plasticity index are slightly different from fill materials, which exhibit clay content ranging from 4 to 30% and a plasticity index from NP to 20%. Nonetheless, the WG soil falls within the lower range of fill materials in Korea, and the properties of the soil samples are similar to the representative earth-fill materials found in the W.A.C. Bennett Dam in Canada, which showed the specific gravity of 2.7, fine content of 20 to 30%, and ${C}_{u}$ of 16 to 36 (Li 2008; Moffat et al. 2011).

Fig. 1

Particle size distribution of WG soil and fill dam materials in South Korea

Table 1 Geotechnical properties of soil sample

Experimental apparatus

The suffusion test is a quantitative internal erosion evaluation method that allows for the measurement of eroded soil within the soil mass, distinct from previous erodibility tests such as the SCS Dispersion Test, Crumb Test, and Pinhole Test (Reddi et al. 2000; Wan 2006). In this study, Aa suffusion test apparatus was developed based on the experiments conducted by the U.S. Army Corps of Engineers (USACE 1953; Kenney et al. 1985; Kenney and Lau 1985; Wan and Fell 2008; Chang and Zhang 2013a). The apparatus was designed to facilitate the easy measurement of eroded soil mass and flow rate during the tests. Therefore, it was possible to evaluate the changes in eroded soil mass and permeability over time, during the tests. This apparatus was also used in the tests conducted by Lee et al. (2021, 2022), and a detailed description of the experimental apparatus was provided by Lee et al. (2022).

The suffusion test apparatus used in this study consisted of four main parts: an inflow water tank, an inner cell, an outer cell, and a replaceable collector (Figs. 2 and 3). The water-supply tank maintained a constant water level at the inlet. The water in the outer cell served to maintain the degree of saturation of the specimen and a constant water level of the outflow during the test or while replacing the collector to measure the amount of eroded soil. A cylindrical specimen (diameter and height of 10 cm) was installed above the perforated plate within the inner cell, and the inner cell was submerged inside the outer cell. The plate had 80 holes with sizes corresponding to the filter criteria (particle diameter at which 15% mass passing through filter material should be smaller than four times that at which 85% mass passing through base material) suggested by USACE (1953). Because the D85 of the WG soil is 2.5 mm, the D15 of the filter material should be less than 10 mm. As shown in Fig. 4, the constriction sizes of 10-mm particles are 1.55 and 4.14 mm in dense and loose cases, respectively (Kezdi 1979; Locke 2001). Therefore, the plate was perforated with holes 4 mm in diameter. As water flowed through the specimen and out of the outlet tube, eroded soil accumulated in the replaceable collector via a funnel connector. The amount of water was measured directly from the outlet tube during the test. The amount of eroded soil was obtained by drying the soil collected from the replaceable collector at predetermined time intervals.

Fig. 2

Schematic design of the experimental apparatus

Fig. 3

Experimental apparatus

Fig. 4

Constriction size of loose and dense particles

Experimental procedure and program

The inner cell was inverted upon sample preparation to prevent segregation and loss of soil during compaction. A porous stone was placed beneath the sample to protect the soil from scouring and to ensure uniform water distribution over the specimen. The soil samples were then compacted to reach predefined relative densities (${D}_{r}$) of 50, 65, and 80%, according to the standard test method for evaluating the permeability of soils, ASTM D2434 (ASTM 2022). The Korean Ministry of Land, Transport and Maritime Affairs (2011) requires that the range of dry unit weight of fill materials in Korea be 90 to 95% of the maximum dry unit weight. The range of dry unit weight of fill materials obtained from 36 samples in nine different Korean fill dams was 1.27 to 1.86 g/cm³ (average 1.63 g/cm³) (Kim 2019). Therefore, the samples in this study were prepared with the target dry unit weight of 1.5, 1.6, and 1.7 g/cm³, which corresponds to the relative density of 50, 65, and 80%, respectively. After the sample preparation, the perforated plate was placed in the inner cell. The cell was rotated back to its original position such that the porous stone and perforated plate were on top of and underneath the specimen, respectively. During the saturation process, the inner cell was submerged in the outer cell for 24 h. Water was introduced across the specimen by adjusting the height of the water tank according to the desired hydraulic gradient and connecting the water tank to the inner cell. The height difference between the water tank and the outlet tube becomes the hydraulic head applied to the specimen. The hydraulic gradient was established based on the short-term test results provided by Lee et al. (2022) to confirm both the internally unstable and stable states of the soils. In previous short-term tests, soils prepared at 50, 60, and 80% relative density showed internal instability under the hydraulic gradient of 35, 60, and 130. Therefore, in the current study, the soils prepared at 50% relative density were subjected to the hydraulic gradients of 25 (internally unstable hydraulic gradient) and 17 (half of the hydraulic gradient of 35), followed by 60 and 30 for 65% relative density and 120 and 60 for 80% relative density. Additionally, hydraulic gradients 5 and 15 were added to well-graded soil with a relative density of 50% to investigate the effect of lower hydraulic gradient (Table 2). Although the usual hydraulic gradient experienced by dams and their foundations is less than 5 (Fell et al. 2015), higher hydraulic gradients have been applied to observe the internal instability due to the time-consuming nature of the experiment. The amount of eroded soil, flow rate, and settlement were measured (for 576 h) under a constant hydraulic gradient. The weights of the eroded soil and the settlements were measured at regular intervals (1, 2, 4, 8, 12, 24, and 48 h). To calculate the coefficient of permeability, the amount of water discharged from the outlet tube was measured at three to four regular intervals between each measurement of the eroded soil. After the tests, a particle size analysis was conducted on the eroded soil and post-test specimens.

Table 2 Experimental program

Results and discussion

The amount of eroded soil, flow rate, and settlement were measured during the tests, and k, e, and porosity (n) were calculated over time. Additionally, the particle size distributions of the eroded soil and post-test specimens were estimated based on sieve analysis. A summary of the test results is presented in Table 3, and the detailed test results are described in the following section.

Table 3 Summary of the test results

Amount of eroded soil and erosion rate

The amount of soil eroded over time is shown in Table 3 and Fig. 5. Based on the post-test particle size distribution analysis, the percentage of fines, i.e., particles smaller than 0.075 mm denoted herein as FS, in the discharged soil, ranged from 48 to 78% (Fig. 6). The test results in Fig. 5 show that the soil erosion rate increases as the hydraulic gradient increases at a constant relative density. This is because a higher hydraulic gradient produces a higher seepage force and flow rate on the soil. Consequently, the soil erosion rate accelerates with an increased hydraulic gradient.

Fig. 5

Accumulated eroded soil of all test results over time

Fig. 6

Particle size distributions of the initial specimen (WG soil) and discharged soil

When comparing the erosion rates (up to approximately 40 h of testing time) based on relative density, the higher the relative density, the lower the soil erosion rate at similar hydraulic gradients (Fig. 7). Because fine particles in WG soils are likely to participate in transferring the effective interparticle stress, the relative density of WG soils will have a more significant effect on resistance to internal erosion. Thus, the higher the relative density of the WG soils, the higher the effective stress that can be applied to fine particles, resulting in improved resistance to internal erosion (Shire et al. 2014). As shown in the test results of WG50-25 and WG65-30 and WG65-60 and WG80-60 in Table 3, increasing the relative density under similar hydraulic gradient conditions interferes finer particles to move, leading to a smaller amount of eroded soil.

Fig. 7

Erosion rate of all test results with respect to hydraulic gradients

Hydraulic conductivity

The results were classified into internally unstable and stable according to the amount of accumulated eroded soil, to analyze the permeability change during the tests. The minimum limit of erosion on unstable soils is defined as 5% (Kenney and Lau 1985; Lee et al. 2022). Based on the predefined soil loss (5%), the test results at hydraulic gradients of 5 and 15 were evaluated as stable. In contrast, those at hydraulic gradients of 17 and 25 were evaluated as unstable for the soil at a relative density of 50%. WG soil (${D}_{r}$ = 65%) at a hydraulic gradient of 30 was evaluated as stable, whereas WG soil (${D}_{r}$ = 65%) at a hydraulic gradient of 60 was evaluated as unstable. Additionally, WG soil (${D}_{r}$ = 80%) was assessed as stable at a hydraulic gradient of 60 and unstable at a hydraulic gradient of 120 (Table 3).

Internally stable test results

The WG soil, classified as internally stable, reduced k and convergence to a constant value despite soil erosion. Figure 8 shows the relationship between the accumulated eroded soil and k normalized by the initial k (${k}_{0}$) at hydraulic gradients of 5 and 15 for WG soil with a 50% relative density. The results show that after a rapid reduction in k, it tends to converge at a lower level while exhibiting small fluctuations within a range less than ${k}_{0}$. The movement of fine particles, clogging, and breakage of the soil structure cause the change in k. Particle size distribution analysis was performed by dividing the sample into upper, middle, and lower parts after the experiment under a hydraulic gradient of 15 to confirm the movement of fine particles and clogging. The FS contents of the top, middle, and bottom sections were 42.0, 42.3, and 44.6, respectively (Fig. 9). In contrast to the initial FS content (42.8%) of the soil, the FS contents of the top and middle parts decreased, whereas that of the bottom part increased, indicating that the fine particles in the top and middle parts moved downward and clogged in the bottom part.

Fig. 8

Accumulated eroded soil (a) and normalized k (b) of WG soil (${D}_{r}$ = 50%) at hydraulic gradients of 5 and 15

Fig. 9

Particle size distribution of WG soil (D_r = 50%) at a hydraulic gradient of 15

To evaluate the effect of FS migration on the change in permeability, particle size distribution analysis and the graphical technique proposed by Kenney and Lau (1985) were used to obtain D₁₀ (particle diameter corresponding to 10% passing by weight) and e, respectively. As demonstrated in Table 4, once the fine particles clogged the bottom part, D₁₀ and e decreased, reducing local permeability. In contrast, they increased in the top and middle parts owing to the selective erosion of the FS contents. It should be noted that a decrease in local permeability can lead to a reduction in overall permeability.

Table 4 D₁₀ and e of WG soil (${D}_{r}$ = 50%) at a hydraulic gradient of 15

To validate the reduction in the overall permeability due to the decrease in the local permeability, the k of each part was calculated using the predictive method (Eq. 1) (Chapuis 2015). The overall k value was calculated using Darcy’s law for continuous flow in saturated media (Eq. 2), where k is the coefficient of hydraulic conductivity; D₁₀ is the diameter of the 10% mass passing; e is the void ratio; v is the velocity of flow; and i is the hydraulic gradient.

$$k\left(cm/s\right)=4.236 {\left(\frac{{D10}^{2}{e}^{3}}{1+e}\right)}^{0.925}$$

(1)

$$v=k\times i$$

(2)

As shown in the results (Table 4), D₁₀ and e increased in the top and middle parts owing to the migration of fine particles, leading to a higher k. Meanwhile, in the bottom part, a reduction in k was observed because the clogging of the fine particles reduced D₁₀ and e. Although D₁₀ and e of the entire specimen increased, the total k decreased. Furthermore, the hydraulic gradient of the bottom part increased by 1.8 times the overall hydraulic gradient. Therefore, it can be concluded that when the overall hydraulic gradient is constant and fine particles are locally clogged within the soil, a decrease in the local permeability can reduce the overall permeability, resulting in a more significant change in local hydraulic gradient. However, it should be noted that the primary purpose of using the Chapuis equation (Eq. 1) is to describe the potential reduction in overall hydraulic conductivity resulting from an increase in fine contents at the bottom. Thus, the equation in this study was primarily employed to address the trend in hydraulic conductivity change rather than providing absolute values.

When k in the bottom part decreases, the pore water pressure in the bottom part increases. The clogging can be partly released, resulting in a temporary increase in k. However, the clogging cannot be completely broken, and k decreases again because of the repeated migration and clogging of the fine particles. Consequently, k gradually decreases after a series of fluctuations. The flow rate decreased owing to the reduction in k. Consequently, the migration of fine particles and clogging were reduced, leading to the convergence of k.

Even in the test results of WG soil at hydraulic gradients of 30 (${D}_{r}$ = 65%) and 60 (${D}_{r}$ = 80%), the convergence of k toward a lower level was accompanied by fluctuations in k (Fig. 10). The FS content of the top and middle parts decreased and that of the bottom part increased compared with the initial FS content of the soil specimen.

Fig. 10

Normalized k of WG soil at ${D}_{r}$ = 65 and 80% for hydraulic gradients of 30 and 60, respectively

Internally unstable test results

WG soil (${D}_{r}$ = 50%) at hydraulic gradients of 17 and 25, which were evaluated to be internally unstable, showed a reduction in k and a gradual decrease in erosion rate until the accumulated eroded soil reached approximately 4% of the total mass of the specimen (Figs. 11 and 12). At 4% or more, however, k and the erosion rate of the soil increased rapidly. Particle size distribution analysis after the seepage experiment at a hydraulic gradient of 17 showed that the FS content of each part was 36.4, 34.8, and 29.1% in the top, middle, and bottom parts, respectively (Fig. 13), indicating that fine particles were reduced in all parts when compared to the initial FS content (42.3%). It should be noted that, unlike the particle size distribution analysis result of the internally stable soil at a hydraulic gradient of 15, the FS content in the bottom part was the lowest. Therefore, it can be deduced that the decrease in k before the onset of internal instability can be attributed to clogging. The subsequent abrupt increase in k results from the breakage of clogging and erosion of fine particles.

Fig. 11

Accumulated eroded soil (a) and k (b) of WG soil (${D}_{r}$ = 50%) at a hydraulic gradient of 17

Fig. 12

Accumulated eroded soil (a) and k (b) of WG soil (${D}_{r}$ = 50%) at a hydraulic gradient of 25

Fig. 13

Particle size distribution of WG soil ($D_{\mathrm r}$ = 50%)  at a hydraulic gradient of 17

The variation in k during the seepage test observed in this study is analogous to that observed in previous experimental studies. Reddi et al. (2000) reported that in gap-graded soil (sand and kaolinite mixture), clogging and washout of fine particles caused a decrease and sudden subsequent increase in k with rapid soil erosion. A similar phenomenon was obtained in well-graded soil, where a reduction in k was followed by a sudden increase in k (Liu et al. 2021).

Similar results were obtained in the test on WG soil (${D}_{r}$ = 65%) at a hydraulic gradient of 60 (Figs. 5 and 14). Initially, k exhibited an irregular pattern. However, over time, k decreased to approximately 4% of the total eroded soil. Subsequently, k and the erosion rate suddenly increased, similar to the internally unstable test result of WG soil at ${D}_{r}$ = 50%. In WG soil (${D}_{r}$ = 80%), at a hydraulic gradient of 120, up to approximately 4% of cumulative eroded soil, k, and the erosion rate of the soil decreased over time (Figs. 5 and 14). When the percentage of eroded soil was 4% or higher, the erosion rate increased rapidly. However, in contrast to the test results at low relative densities, k did not increase dramatically, even though the rate of soil erosion increased significantly. This difference was due to the settlement of the specimen, which resulted in a decrease in e. A detailed discussion of this discrepancy is presented in the following section.

Fig. 14

k of WG soil at $D_{\mathrm r}$ = 65 and 80% for hydraulic gradients of 60 and 120, respectively

Settlement

Internal instability was accompanied by settlement in the WG soil (Fig. 15). For all test cases, the relationship between the amount of eroded soil and settlement showed a significant difference between the internally stable and unstable soils. There is no clear relationship between the settlement and amount of eroded soil in internally stable soils. However, in internally unstable soils, the settlement increases as the amount of eroded soil increases. As shown in Figs. 5 and 16, the settlement patterns and cumulative eroded soil over time were very similar, indicating that soil erosion was accompanied by settlement. Therefore, it can be inferred that the erosion of WG soil occurs along with the deformation or collapse of the soil structure, unlike in gap-graded soil. The deformation of the soil structure can be partially proven by the coarse particles within the eroded soil (the percentage of coarse particles larger than 0.075 mm in eroded soil ranged from 22 to 35%), which are the main elements of the soil structure (Fig. 6). This results from the fact that during soil erosion, not only the fine particles but also the coarse particles flowed out, resulting in suffosion, in which the soil structure collapsed, and the volume decreased.

Fig. 15

Accumulated eroded soil and settlement of test results

Fig. 16

Settlement of WG soil at ${D}_{r}$ = 50, 65, and 60% for hydraulic gradients of 17, 60, and 120, respectively

For the internally unstable test results, the variation in e was calculated based on the volume change due to the observed settlement and the amount of eroded soil. For WG soil (${D}_{r}$ = 50%) at a hydraulic gradient of 17 and WG soil (${D}_{r}$ = 65%) at a hydraulic gradient of 60, e increased by 0.05 and 0.11, respectively, despite the settlement, increasing k indicating that the influence of the eroded soil was more dominant (Figs. 11 and 17, and Figs. 14 and 18). In the case of WG soil (${D}_{r}$ = 80%) at a hydraulic gradient of 120, although 6.7% of the soil was discharged, a relatively higher settlement occurred, leading to an increase in e only by 0.01 (Figs. 14 and 18).

Fig. 17

e of WG soil (${D}_{r}$ = 50%) at a hydraulic gradient of 17

Fig. 18

e of WG soil at ${D}_{r}$ = 65 and 80% for hydraulic gradients of 60 and 120, respectively

In gap-graded soils, because fine particles can quickly move through the constrictions between the coarse particles supporting the soil structure, suffusion can occur without changing the overall volume of the soil (Lee et al. 2021, 2022). However, fine particles may transfer effective interparticle stress in the WG soils. Therefore, when the seepage force erodes fine particles, the soil structure collapses, inducing settlement (Sato and Kuwano 2015). Moreover, as the relative density of the WG soil increased, the finer particles were under higher effective stress, requiring a higher hydraulic gradient to cause internal instability. When the hydraulic gradient is high, rearrangement and redeposit of coarse particles occur (Reddi et al. 2000) together with the erosion in fine particles, resulting in significant settlement. Consequently, the WG soils with higher relative density can exhibit greater settlement under a hydraulic gradient which can induce internal instability.

Mechanism of internal instability

Based on the test results and analysis, the mechanism of internal instability and its progress in the WG soil can be summarized as follows: The internal instability steps are numbered sequentially and comprehensively compared with the test results (Fig. 19).

Fig. 19

Schematic diagram of internal instability progression in WG soil

At the beginning of the tests, some fine particles flowed out from the bottom part of the specimen, mainly owing to gravity (Step 1). Consequently, k exhibits small fluctuations within a range less than k₀ (Figs. 8 and 10). When the fine particles begin to move because of the seepage force, some fine particles become clogged, reducing k, and some fine particles are discharged (Step 2). Consequently, a local reduction in k gradually decreases the flow rate, resulting in a reduction in soil erosion and an internally stable state (Fig. 20). During this step, no significant settlement developed, and internal stability was maintained (Fig. 15).

Fig. 20

Internal instability behavior in WG soil

The clogging was resolved when a sufficient seepage force accumulated or a higher seepage force was applied. The fine particles clogged at the bottom of the specimen started to flow out together with the coarse particles (Step 3). Because the effect of the selective erosion of fines was more dominant during this process, D₁₀, e, and k tended to increase (Figs. 11 and 12). The migration of fine particles was followed by coarse particle discharge, soil structure collapses, and settlement (Fig. 16), leading to soil structure rearrangement (Step 4). Therefore, the internal instability of the WG soil occurred in the form of suffosion (Fig. 20).

Conclusions

In the previous short-term test results of the WG soil, which demonstrated the representative particle size distribution for fill dam soils in South Korea, it was proposed that the characteristics of the internal instability and its development process of the WG soil were different from those of the gap-graded soil (Lee et al. 2022). Therefore, in this study, long-term suffusion tests were conducted with various relative densities and constant hydraulic gradients to analyze the progress and cause of the internal instability of the WG soil. Based on the test results, the specific findings of this study are summarized as follows.

The higher the hydraulic gradient, the faster the erosion rate of the soil, owing to the greater seepage force acting on the soil. In addition, when the relative density increased, the amount of eroded soil decreased for the same hydraulic gradient. The fine particles in the WG soil may transfer effective interparticle stress. Thus, the higher the relative density of the WG soils, the higher the effective stress that can be applied to fine particles, resulting in improved resistance to internal erosion.

Under internally stable conditions, the WG soil showed a reduction in k and converged to a constant value despite soil erosion. Based on the particle size distribution analysis, the fine particles in the upper and middle parts moved downward and clogged the lower part of the specimen. Once the fine particles clogged the lower part, D₁₀ and e decreased, reducing the overall permeability. The flow rate decreased owing to the decrease in k; consequently, the migration of fine particles and the amount of clogging decreased, leading to k convergence.

WG soil, which was evaluated to be internally unstable, showed a k and erosion rate reduction until the cumulative eroded soil reached approximately 4%. However, when the amount of accumulated eroded soil reached 4% or more, k and the erosion rate of the soil increased rapidly with a reduction in the FS content in all parts of the specimen. From the test results, it can be concluded that D₁₀ and e increased as clogging was released and fine particles were discharged, resulting in an increase in k.

In the internally unstable test results, settlement increased as the amount of eroded soil increased. The erosion of fine particles results in the collapse of the soil structure and migration of coarse particles, which leads to internal instability in the form of suffosion. When internal instability occurs at low relative densities, k increases significantly because the effect of the selective erosion of fine particles is more dominant than that of settlement. However, as the relative density of the soil increased, the effect of the settlement also increased, resulting in no significant change in k.

Based on the test results and analysis, the mechanism of internal instability and its progression in well-graded soils were proposed. In WG soils, the internal instability proceeds in the form of suffosion, which is distinct from the suffusion mechanism in gap-graded soils. While long-term prediction of internal instability in hydraulic earth structures remains challenging, our findings provide valuable data to enhance our understanding of the long-term development of internal instability in well-graded soils. However, due to the application of a relatively higher hydraulic gradient in this study, additional validation tests are necessary to extend the mechanism to actual fill dam conditions. This could involve conducting longer experiments using actual fill materials under hydraulic gradients encountered in actual fill dams. Additionally, investigating the movement path of fine particles by calculating the k at various depths using the pore water pressure transducers would be beneficial.