Experimental Characterization of the Seismic Responses of Silt-Interlayered Stratified Sandy Tailings Slopes Using Centrifuge Shaking Table Tests
This study aims to characterize the seismic responses of silt-interlayered stratified sandy tailings slopes. In order to achieve this, a series of model tests were carried out in an advanced centrifuge shaking table. The experimental results demonstrate that sandy tailings slopes interlayered with silts are more prone to flow failure compared with homogeneous slopes. The failure process of a steep silt-interlayered sandy slope is further elaborated in detail, and two modes of tailings release have been identified. The initial loading cycles may only lead to localized deformation near the relatively shallow silty layer; however, there is continuous build-up of excess pore water pressure in deeper sandy tailings. With increasing the number of loading cycles, more sandy tailings start to spread laterally, and ejection of tailings may occur. After ejection, the excess pore pressure turns to dissipate, while the tailings near slope crest become liquefied and then flow toward the downstream direction layer by layer, leading to tailings release. This is the first mode of tailings release. Due to the continuous lateral spreading, the displaced tailings eventually arrive at the crest of the starter dam and then flow along its slope at the downstream side, leading to tailings release. This is the second mode of tailings release. Those unique and invaluable observations can be used to improve the seismic design of tailings impoundments and to validate associated numerical methods.
Tailings dams are often used to accommodate tailings [1–5]. Compared with water retention dams, such tailings dams are more prone to failure based on statistical data [6, 7]. Tailings dam failures can result in high-velocity long-run-out mudflows, leading to immediate fatalities, direct economic losses, and long-term impacts on the ecosystem [8–10]. As summarized by Santamarina et al. , nearly 3000 people have been killed and nearly 270 mm3 tailings have been released in past failure events. Seismic loading is one of the main triggers of those failures . Table 1 summarizes the case histories of earthquake-induced failures, e.g., the El Cobre no. 1 dam failure which caused release of 1.9 mm3 tailings and 300 fatalities [14, 17]. Therefore, it is important to examine the seismic responses of tailings dams.
Unlike conventional earth dams founded on sufficiently strong grounds, tailings dams are often constructed using the upstream method with sub-dams being founded on saturated tailings. Those tailings may liquefy or suffer a major reduction in shearing resistance under seismic loading [11, 15, 18, 19]; that is why most failures occurred in upstream dams as illustrated in Table 1. Therefore, the stability of tailings dams is mainly determined by the properties of the underlying tailings rather than the materials in sub-dams. The tailings deposits may contain various patterns of layering and lensing due to the deposition of particles through water. Silt-interlayered stratified sandy tailings are often encountered in the field, e.g., the Mochikoshi tailings dam in Japan  and the Lixingou tailings dam in China . Based on the 1 g soil column tests and slope model tests on sands with similar soil structures performed by Kokusho , water may accumulate under the weakly permeable layer to form a water film with a thickness of a few millimeters, which imposes a strong influence on the seismic behavior. This highlights the challenge in analyzing the seismic responses of silt-interlayered sandy tailings slopes .
As previously mentioned, the tailings slope failures are associated with large deformations, liquefaction, detachment, and flow of tailings. Such complicated behavior can hardly be simulated using numerical methods based on various pore pressure generation models, e.g., the Finn model [23, 24], the UBCSAND model [25–27], and the CycLiq model . Shaking table tests at 1 g (gravitational acceleration) condition have been used to examine the seismic behavior of tailings grounds or slopes. However, homogeneous conditions are used in nearly all those tests, e.g., [29–31]; hence, the test results can hardly reflect the behavior of a stratified slope. Moreover, because of the limited sample size and low effective stresses of those 1 g model tests, the observed seismic responses are not easily transposed to in-situ behavior during actual earthquakes [30, 32, 33]. Because of those reasons, the 1 g shaking table tests can hardly simulate the phenomenon observed in the field, e.g., sand boiling , mudflows along the downstream slope of the starter dam [15, 16], and the topographic features, e.g., terraces and scarps of the tailings left in place after failure [12–14]. As the in situ stress field can be reproduced in the centrifuge, centrifuge shaking table tests are believed to be more efficient and advantageous [34–38]. However, there is lack of such kind of experimental results.
This study aims to characterize the behavior of silt-interlayered stratified sandy tailings slopes subjected to seismic sequences. A series of model tests have been carried out in an advanced centrifuge shaking table. The experimental details are firstly presented, followed by experimental results and discussions. The focus is put on the deformation behavior and how the tailings are released from the impoundments. The salient findings can shed light on the associated failure mechanisms and then improve the seismic design of tailings dams. The unique and invaluable observations can also be used to validate constitutive models, procedures, and results of numerical methods.
2. Experimental Details
The model tests were carried out in a centrifuge shaking table at China Institute of Water Resources and Hydropower Research (IWHR). As shown in Figure 1, the beam centrifuge has a capacity of 450 g ton and a radius of 5.03 m. The centrifuge shaking table can operate up to 100 g with a payload mass of up to 440 kg. It can simultaneously simulate horizontal and vertical bedrock motions with peak accelerations of up to 30 g and 20 g at model scale, respectively. This study only used horizontal dynamic excitation, and the centrifugal acceleration was set to 40 g, i.e., N = 40. As summarized in Table 2, the scaling factors follow the well-established scaling laws for centrifuge shaking table tests . In order to avoid the time scaling conflict for consolidation and for the seismic motion, the viscosity of the pore fluid was increased 40-fold by using an aqueous solution of hydroxypropyl methylcellulose (HPMC), which is commonly used in centrifuge shaking table tests (e.g., ).
2.1. Model Configuration
As summarized in Table 3, three centrifuge shaking table tests were carried out. Each model was prepared in a rigid container with inner dimensions of 0.9 m (length) × 0.2 m (width) × 0.5 m (height). The container was equipped with a transparent Perspex side window to facilitate the image-based analysis (see Figure 1). As shown in Figure 2, each model was mainly composed of a starter dam and a tailings slope. Each starter dam has a height of 160 mm and a slope ratio of 1 : 1.5 (height: width), and each tailings slope has a height of 260 mm. The tailings slope in test T1 was constructed with homogeneous sandy tailings, while silt-interlayered stratified sandy tailings were used in both tests T2 and T3. Two silty layers were sandwiched in the sandy tailings at heights of 230 mm and 180 mm, i.e., at 88% and 69% of the total model height, respectively. The slopes in tests T1 and T2 had a ratio of 1 : 5, while a steeper slope with a ratio of 1 : 3 was adopted in test T3. The range of the slope ratios has been frequently used in the tailings dams in both China (e.g., ) and the rest of the world (e.g., [12, 16]). Since the sizes of sub-dams were relatively small compared with the whole tailings slope in real projects, those sub-dams were not separately constructed in the models to facilitate model preparation. This treatment had a minor effect on the seismic responses of the whole tailings slopes and was commonly used in previous model tests (e.g., ).
2.2. Model Preparation
For each model, the starter dam was firstly prepared using dry coarse silica sands with particle sizes of 2–4 mm produced in Fujian Province, China. The sands were densified to 1.9 g/cm3, which was close to the value used in practice. Afterward, as shown in Figure 3(a), the dry tailings slope was prepared using tailings collected from a real tailings impoundment of a gold mine in Gansu Province, China. The tailings had a maximum size of 0.1 mm and contained 31.2% of fines with sizes smaller than 0.075 mm; hence, it was categorized as silty sand based on the Unified Soil Classification System . The air pluviation method was used to prepare the sandy tailings slope, and the achieved dry density was 1.40 g/cm3, corresponding to a relative density of 66%. Regarding the silty layers used in tests T2 and T3, dry silts were gently placed on the surface of the sandy tailings at the prescribed elevation to form a layer with a thickness of 5 mm. The density of the silts was 1.44 g/cm3. The effective particle size d10 of the silts was 1∼10 μm, which was close to the field value (e.g., ). As the hydraulic conductivity was mainly determined by d10, this ensures the similarity of soil properties of prototype and model silty layers. In practice, the phreatic surface was only slightly lower than the slope surface, and there was steady seepage in saturated soils beneath the phreatic surface . Considering the difficulty in maintaining the seepage steady in the centrifuge, the seepage field was not reproduced in this study. The centrifuge model was immersed in liquid to reproduce the saturation condition of tailings. The effect of the initial seepage force before shaking on the seismic responses of tailings slopes was not considered in this study.
2.3. Instrumentation Layout
As shown in Figure 2(a), series of vertically distributed miniature pore pressure sensors were installed at various depths in the upstream part to monitor the generation and dissipation of pore pressure in each test. In addition, two high-speed cameras (model: Hero 8, GoPro Inc., USA) and particle image velocimetry (PIV) techniques were used to capture the displacement field. As shown in Figure 1, the cameras were installed on the inner side walls of the test basket of the centrifuge. The spin-up process was monitored at a frame rate of 30 fps and a resolution of 4000 × 3000 pixels, while a frame rate of 240 fps and a resolution of 1280 × 960 pixels were used during shaking. As shown in Figure 3, since the texture of tailings could hardly be identified, trace markers made by silica powders with a size of 48 μm were distributed at several horizontally distributed arrays, e.g., arrays 1, 2, and 3, to visualize motion.
2.4. Experimental Procedures
Each dry model was saturated under vacuum using an aqueous solution of HPMC in a saturation box. The mass flux of the aqueous solution was controlled to as low as 0.3–0.4 kg/h using a peristaltic pump to prevent fluidization of the soils. The saturated model was then mounted on the centrifuge shaking table (see Figure 1). After completing all those preparation steps, the centrifuge started to spin up and the centrifugal acceleration gradually increased to 40 g. The model was subjected to horizontal seismic motion after a specific time of stabilization. As shown in Figure 2, the shaking direction was along the longitudinal direction of the container and was perpendicular to the dam axis.
In order to facilitate monitoring the seismic responses during shaking, the long-duration small-magnitude seismic sequence rather than an input motion with shorter duration and larger magnitudes was used in each test. Figure 4 presents the recorded input motion in both the time and frequency domains in test T3. Similar measurements were also derived in tests T1 and T2. The total duration of the seismic sequence was 120 s, corresponding to 80 min at prototype scale. The peak acceleration in each cycle was 1–7.2 g at model scale, corresponding to 0.025–0.18 g at prototype scale. The input motion had high frequency components at 36 Hz, corresponding to 0.9 Hz at prototype scale. The total number of loading cycles was estimated to be in the range of 4000–5000. Therefore, each model was subjected to a seismic sequence of hundreds of small-magnitude earthquakes in a relatively short period, which roughly simulated the real earthquake sequences (e.g., [43–45]).
3. Results and Discussion: Effects of Slope Ratio and Soil Structure
The postearthquake slope conditions are firstly compared with each other. As shown in Figure 5(a), the homogeneous tailings slope is stable and only settles for about 7 mm at model scale or 28 cm at prototype scale, which accounts for about 3% of its total height. Although the same slope ratio is used in test T2, the tailings near slope crest have been released, and the total height is reduced to about 91% of its initial value (see Figure 5(b)). In addition, the sandy tailings between the two silty layers spread laterally with a maximum horizontal displacement of about 20 mm at model scale or 80 cm at prototype scale. The comparison between the results from tests T1 and T2 highlights the effect of soil structure on the seismic behavior of tailings slopes. With regard to the relatively steep silt-interlayered slope in test T3, much more tailings have been released, and the total height is reduced to about 83% of its initial value (see Figure 5(c)). The volume of released tailings is estimated to be about 20% of the total value of tailings initially located above the starter dam crest. The comparison between the results from tests T2 and T3 demonstrates the effect of slope ratio.
The experimental results from test T3 are then used to elaborate the failure process of tailings slopes subjected to earthquake sequences. The whole process has been roughly divided into four stages according to the seismic responses in terms of soil deformation to facilitate the discussion. In stage 1, deformation is localized in the relatively shallow silty layer and the sandy tailings near slope crest. In stage 2, soil movement occurs at deeper locations. In stage 3, sand boiling occurs and the tailings start to be released. In stage 4, a new tailings release mode is observed. As shown in Figure 3(c), a coordinate system is used to indicate the initial locations of trace markers. The origin is put at the initial slope crest, while x0 and z0 denote horizontal and vertical distances from the origin to a particular point. The discussions on soil deformation are mainly based on the results of array 1 near upper surface and arrays 2 and 3 between the two silty layers. The average values of z0, i.e., initial depths of arrays 1, 2, and 3, are 4 mm, 40 mm, and 60 mm at model scale, respectively. The following discusses the seismic responses stage by stage, and all results are presented in model units unless otherwise indicated.
4. Results and Discussion: Behavior in Stage 1
The initial loading cycles only lead to localized deformation near the relatively shallow silty layer. The duration of this stage is about 8 s.
4.1. Soil Deformation
Figure 6 shows a comparison of the slope profile before shaking and that at t = 8 s. This comparison suggests lateral spreading of the relatively shallow silty layer and the sandy tailings near slope crest. Figures 7(a) and 7(b) show both horizontal displacement dh and vertical displacement of arrays 1 and 2 (see Figure 3(c) for the locations of the arrays), respectively. As illustrated by the snapshot in Figure 7(a), a positive dh value indicates that the trace marker moves toward the downstream direction, while a negative value represents movement in the opposite direction. A positive value of indicates downward movement (see the snapshot in Figure 7(b)), and a negative value, upward movement. The trace markers in array 1 continue to settle with increasing the number of loading cycles. At the end of stage 1, the settlement is 2.8–3.5 mm at x0 = 35–155 mm where the gravitational driving force is limited. This value is comparable with the initial thickness of the relatively shallow silty layer, which is ∼4 mm before shaking. The underlying trace markers at x0 = 57–147 mm in array 2 only settle for less than 0.4 mm at t = 8 s, which only contributes to a small part of the settlement of the trace markers in array 1. Therefore, the settlement of trace markers in array 1 should be mainly attributed to the reduction of the thickness of the relatively shallow silty layer caused by lateral spreading. Such lateral spreading also leads to horizontal movement toward the downstream direction in the upper sandy tailings (see the results from array 1 in Figure 7(a)). It is worth noting that the soils at x0 > 120 mm in array 1 move toward the upstream direction as the profile at this region is not perfect level.
As shown in Figure 7(a), the horizontal displacements at x0 = −128 to −4 mm in array 2 are 1-2 mm at t = 8 s, while the tailings at x0 > 57 mm in the upstream part exhibit horizontal displacements less than 0.4 mm. For the trace markers at array 3 placed at deeper locations, only those at x0 = −144 to −99 mm close to the slope surface exhibit values of dh that exceed 1 mm. This suggests a localized deformation below the downstream end of the relatively shallow silty layer, which is also reflected by comparing the slope profile at t = 0 and that at t = 8 s (see Figure 6).
4.2. Generation of Excess Pore Pressure
Figure 8(a) presents the evolutions of excess pore pressure Δu at various depths in the upstream part over the whole shaking process in test T3. The values of Δu at all depths increase over time in stage 1, suggesting continuous excess pore pressure generation. The evolution of excess pore pressure Δu in test T1 is given in Figure 8(b) to make a comparison. The increase in Δu seems localized at extremely shallow depths, e.g., at P1_T1. As for the tailings at relatively large depths, e.g., at P2_T1 and P3_T1, although an increasing trend of Δu can be identified, the total increment is negligible. This suggests the effect of soil structure on the pore pressure evolution.
The excess pore pressure ratio Ru in test T3 is further calculated by dividing Δu by the initial effective vertical stress. Since the total stress at each sensor location has significantly changed after 30 s because of soil movement, only evolutions of Ru before 30 s are shown in Figure 9. The ratio Ru at P2_T3 below the relatively shallow silty layer increases sharply with time and exceeds 0.7 after 2 s. This is similar to the observations in the 1 g soil column tests and slope model tests on loose sands with weakly permeable thin silty interlayers carried out by Kokusho [21, 46]. Due to the relatively weak permeability of silty layers, the sandy tailings beneath those silts may experience a sharp increase of excess pore pressure with increasing number of loading cycles. The high values of Ru may lead to degradation in soil stiffness and strength, imposing a strong influence on the seismic behavior, e.g., the soil deformation beneath those silty layers.
5. Results and Discussion: Behavior in Stage 2
The subsequent loading cycles in stage 2 from ∼8 s to ∼16 s lead to lateral spreading in a larger zone. Although detachment and translocations of tailings may occur, the tailings release is negligible.
5.1. Lateral Spreading
Figures 10 and 11 show the displacements of trace markers in arrays 2 and 3 between both silty layers (see Figure 3(c) for the locations of arrays), respectively. The horizontal displacements far exceed the vertical displacements in both arrays, suggesting lateral spreading of sandy tailings. The horizontal displacement decreases with increasing x0, and there is minor displacement beyond a threshold value of x0. For example, the values of dh are 0.8–2.1 mm at x0 < 16 mm at 10 s, while the values of dh are less than 0.4 mm at x0 > 57 mm. The threshold value of x0 ranges within 16–57 mm at 10 s for array 2. The threshold distance becomes 131–147 mm at t = 16 s, suggesting the extension of the lateral spreading zone toward the upstream direction. The same behavior is also observed in array 3, where the threshold distance increases from 8–26 mm at t = 10 s to 88–105 mm at t = 16 s. At the end of stage 2, the width of the lateral spreading tailings is estimated to exceed 230 mm, which is about 92% of the initial model height.
5.2. Detachment and Translocation of Tailings
As shown in Figure 12(a), a protrusion of tailings is formed at the end of stage 1 due to the lateral spreading of tailings beneath the downstream end of the relatively shallow silty layer. Such a protrusion is prone to slide under seismic loading. As shown in Figures 12(b)–12(e), the tailings in the protrusion are gradually detached from the tailings slope and then move along the slope surface. This movement may also promote the lateral spreading of tailings above the relatively shallow silty layer. A new profile containing a terrace and a scarp emerges near the slope crest due to such emplacement. Nevertheless, the tailings release is negligible.
6. Results and Discussion: Behavior in Stage 3
The subsequent loading cycles in stage 3 from ∼16 s to ∼84 s lead to sand boiling, tailings sliding, and liquefaction at extremely shallow depths. Significant release of tailings occurs in this stage.
6.1. Sand Boiling
As shown in Figure 13(a), ejection of tailings or sand boiling occurs at x0 = 250 mm and lasts nearly 1 second from 16 to 17 s. This suggests that the tailings beneath silty layers might reach the liquefaction state, and the liquefied tailings are prone to ejection onto the surface to form sand boils under the high hydraulic gradient. This is consistent with the fact that many sand volcanos have been observed at sites of earthquake-induced failures, e.g., Case 10 in Table 1 . After ejection, as shown in Figure 8, the excess pore pressure turns to dissipate with increasing the number of loading cycles, which is different from the response observed in test T1 where the excess pore pressure remains nearly constant after an initial build-up. This indicates that the ejection of tailings might change the drainage conditions of the tailings and then influence the pore pressure evolution. It is worth noting that sand boiling was not observed in tests T1 and T2.
6.2. Tailings Sliding
Figures 14 and 15 present horizontal distributions of displacements from 16 to 22 s at arrays 2 and 3 between both silty layers (see Figure 3(c) for the locations of arrays), respectively. As shown in Figure 13, some particles in the silty layer may move upward through the interface between tailings and the side window, thus hindering the identification of trace markers. Therefore, the associated distributions after t = 22 s are not given in Figures 14 and 15. Such movement of silts may slightly increase the settlement of tailings which are initially founded on those particles. Nevertheless, the figures indicate that the horizontal displacements far exceed the vertical displacements in both arrays, and the maximum value of dh is observed near slope surface, which follows the trend of lateral spreading observed in stage 2.
Figure 16 presents evolutions of horizontal displacements of the trace markers which are not covered by the silty particles. As indicated by the legend, those trace markers are located at x0 = −144 to −12 mm in the downstream side and at x0 = 201–256 mm in the upstream side. After ejection of tailings, the trace markers in the downstream side move toward the downstream direction in a larger speed. The horizontal displacements of all the trace markers located in the downstream side reach 20 mm at model scale or 0.8 m at prototype scale after a duration of 40–50 s. With regard to the upstream side, the trace markers start to move toward the downstream direction after ejection, and then the values of dh reach a plateau at about t = 25 s, suggesting that horizontal movement has already ceased in those tailings.
The relative displacements of those trace markers are further calculated to investigate soil deformations at different locations. The relative displacement in the downstream side is given by the difference between the displacement at x0 = −144 mm and that at x0 = −12 mm, i.e., dh,−144mm − dh,−12mm. As shown in Figure 17, the relative displacement in the downstream side is only 0.5 mm at t = 35 s, and the value increases to 3.3 mm at t = 50 s. The difference between displacements at x0 = −12 mm and x0 = 201 mm, i.e., dh,−12mm − dh,201mm, is used to reflect the relative displacement in the upstream side. The value continuously increases with time and reaches 14.4 mm at 50 s, which is much larger than that in the downstream side. The average horizontal strain can be roughly estimated by normalizing the relative displacement by the initial distance between the two trace markers. The average strains in the downstream and upstream sides are 2.5% and 6.8% at 50 s, respectively. This suggests that the tailings in the downstream side move as a whole and slide along a failure plane in the relatively deep silty layer or along the interface between the two materials.
6.3. Tailings Release
As shown in Figure 13(b), due to the extremely small confining stress at extremely shallow depths near upper surface, tailings can be easily liquefied under seismic loading. The liquefied tailings then flow along the slope surface and the downstream slope of the starter dam, and eventually arrive at the deposition space between the slope of the starter dam and the side walls of the container. As the flow moves in extremely thin layers, it cannot be clearly identified in the slope surface via the images. However, the accumulation of tailings in the deposition space clearly demonstrates the occurrence of tailings flow (see Figure 13(b)).
As shown in Figures 13(c) and 13(d), because of the continuous flow of liquefied tailings at extremely shallow depths, the slope profile keeps changing over time. An extremely gentle slope with an angle of 1-2° emerges at t = 50 s. Afterward, as shown in Figure 13(e)–13(g), the tailings at extremely shallow depths seem to be liquefied and then flow toward the downstream direction layer by layer. The height of the remaining tailings continues to decrease due to the tailings flow. The collected images indicate that the model height at x0 = 0 is decreased by 14% at the end of stage 3, suggesting a significant release of tailings. Such behavior can also be assessed by the measurements from P1_T3. The sensor P1_T3, initially located at a depth of ∼19 mm, might gradually emerge with the flow of upper tailings. Then, the value reflects the static liquid pressure at the current upper surface. It is worth noting that this sensor was observed on the surface at the end of the test. As shown in Figure 8, the pressure at P1_T3 increases over time from 40 s to 70 s with a total increase of 6.2 kPa, which corresponds to a model height reduction of 16 mm, i.e., ∼6% of the initial model height. It is worth noting that the surface of the tailings slopes was not prepared using compacted coarser sands, which was slightly different from the in-situ condition. This treatment might increase the surface flow of tailings during shaking.
7. Results and Discussion: Behavior in Stage 4
The subsequent loading cycles lead to an increase in the release rate in stage 4 from 84 to 120 s.
7.1. Tailings Release
Because of the continuous tailings sliding in stage 3, the downstream end of those tailings gradually moves downward and along the slope surface. As shown in Figure 13(g), tailings arrive at the crest of the starter dam at t = 84 s and then flow along the downstream slope of starter dam, which resembles field observations [15, 16]. As shown in Figure 18(a), not only the tailings at extremely shallow depths, but also those at deeper locations, flow toward the downstream direction along the slope surface. According to the deposition volume estimated via images, the flow rate can be roughly determined. The average flow rate from 60 s to 84 s in stage 3 was about 6.7 cm3/s, while the average flow rate from 84 s to 120 s in stage 4 is about 16.7 cm3/s, which is about 2.5-fold of the value in previous stage. This suggests that the flow of tailings above the relatively deep silty layer contributes to a significant increase in the release rate of tailings.
7.2. Topographic Feature after Release
As shown in Figure 18, the movement of tailings seems to have ceased in the upstream part, while the downstream part continues to spread laterally. The tailings left in place exhibit a failure scarp at x0 = 50 mm and a 378-mm wide terrace at the downstream side at t = 120 s. Several small terraces and scarps can also be identified at the upstream side. This topographic feature is similar to that observed in real tailings impoundments after earthquake-induced failures, e.g., Barahona no. 1 dam, Chile (see Case 1 in Table 1 ), and El Cobre Viejo dam, Chile (see Case 2 in Table 1 [12, 14]). The failure mechanisms observed in this study can be used to explain this topographic feature. Due to the liquefaction and flow of tailings at extremely shallow depths, the height of the tailings slope decreases with increasing the number of loading cycles. The dissipation of pore pressure in the upstream side makes the associated tailings more resistant to seismic loading, while the tailings close to the slope surface slide along a failure plane in the silty layer or the interface. This eventually leads to the formation of terraces and scarps for the tailings left in place.
8. Concluding Remarks
In this study, a series of centrifuge tests on tailings slopes with various slope ratios and stratification features were carried out. Those tests reproduced the phenomena observed in real tailings impoundments including sand boiling, mudflows along the starter dam slope, and the topographic features of tailings left in place. This indicates the superiority of centrifuge shaking table tests. The experimental results highlight the effects of soil structure and slope ratio on the seismic responses, and the focus is put on the deformation behavior and how the tailings are released for silt-interlayered stratified sandy tailings slopes. The salient findings are summarized in the following.
Compared with a homogeneous sandy tailings slope, a sandy tailings slope interlayered with silts is more prone to flow failure and tailings release under seismic loading, and the released volume increases with increasing slope ratio. The detailed deformation process of a relatively steep silt-interlayered sandy tailings slope is as follows. The initial loading cycles may only lead to localized deformation near the relatively shallow silty layers, followed by lateral spreading of tailings between the silty layers at deeper locations. Those localized deformations can be attributed to the rapid build-up of excess pore pressure. Although detachment and translocation of tailings protrusions may occur at the slope surface, the overall deformation of the tailings slope is minor, and the tailings release is limited. The continuous build-up of pore pressure may lead to ejection of tailings during the subsequent loading cycles. After that, the excess pore pressure in the upstream side turns to dissipate, while the tailings near slope crest become liquefied and flow along the slope surface. The slope profile keeps changing with increasing the number of loading cycles until an extremely gentle slope with an angle of 1-2° emerges. After that, the tailings at extremely shallow depths seem to be liquefied and then flow toward the downstream direction layer by layer, leading to tailings release. In addition, due to the continuous lateral spreading or tailings sliding at deeper locations toward the downstream direction, the tailings may eventually arrive at the starter dam crest and then be released, which contributes to a significant increase in the release rate. In summary, the tailings in a sandy slope with interlayered silts may be released in two modes. The first mode is associated with liquefaction and flow of tailings at extremely shallow depths. The second mode is associated with lateral spreading, tailings sliding, and overtopping starter dam. The seismic responses of tailings dams are highly affected by the geometrical and material properties and the characteristics of the input motions, e.g., the frequency content and duration. Since only limited conditions were tested in this study, further experimental study is required to fully understand the seismic responses of tailings dams and to draw universal conclusions. Moreover, considering the prototype height of the tailings slopes tested herein is about 10 m, the seismic behavior of much higher tailings slopes might be different from the observations in this study. Nevertheless, the unique and invaluable observations can also be used to validate constitutive models, procedures, and results of numerical methods.
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
This research was supported by Open Research Fund Program of State Key Laboratory of Hydroscience and Engineering (no. sklhse-2021-D-05), Key Research and Development Projects of Tibet Autonomous Region (no. XZ202101ZY0002G), National Natural Science Foundation of China (nos. 51809290 and 52078212), the IWHR Research and Development Support Program (no. GE0145B032021), and the Open Research Fund of State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin, China Institute of Water Resources and Hydropower Research (no. IWHR-SKL-202003).
B. Lottermoser, Mine wastes: characterization, treatment and environmental impacts, Springer, Berlin, Heidelberg, Germany, 2007.
S. Azam and Q. Li, “Tailings dam failures: a review of the last one hundred years,” Geotechnical News, vol. 28, no. 4, pp. 50–54, 2010.View at: Google Scholar
M. P. Davies, “Impounded mine tailings: what are the failures telling us?” CIM Bulletin, vol. 94, pp. 53–59, 2001.View at: Google Scholar
J. H. Troncoso, A. Vergara, and A. Avendaño, “Seismic failure of Barahona tailings dam,” in Proceedings of the 3rd Conference of the International Conference on Case Histories in Geotechnical Engineering, Missouri, USA, June 1993.View at: Google Scholar
K. Ishihara, “Post-earthquake failure of a tailings dam due to liquefaction of pond deposit,” in Proceedings of the International Conference on Case Histories in Geotechnical Engineering, Tokyo, Japan, May 1984.View at: Google Scholar
Y. Huang, H. L. Cheng, T. Osada, A. Hosoya, and F. Zhang, “Mechanical behavior of clean sand at low confining pressure: verification with element and model tests,” Journal of Geotechnical and Geoenvironmental Engineering, vol. 141, no. 8, Article ID 06015005, 2015.View at: Publisher Site | Google Scholar
J. Zhang, “Study on tailing engineering geological character and stability of Li-West ditch tailing,” Changan University, Xi'an, China, 2006, Ph.D. Thesis.View at: Google Scholar
P. M. Byrne, “A cyclic shear-volume coupling and pore-pressure model for sand in proceedings,” in Proceedings of the Second International Conference on Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, pp. 47–55, St. Louis, Missouri, March 1991.View at: Google Scholar
P. M. Byrne, D. Roy, R. G. Campanella, and J. Hughes, “Predicting liquefaction response of granular soils from pressuremeter tests,” Static and Dynamic Properties of Gravelly Soils, ASCE, Virginia, USA, 1995.View at: Google Scholar
M. H. Beaty and P. M. Byrne, “UBCSAND constitutive model Version 904aR,” Itasca UDM Web Site, vol. 69, 2011.View at: Google Scholar
M. A. Zardari, H. Mattsson, S. Knutsson, M. S. Khalid, M. V. S. Ask, and B. Lund, “Numerical analyses of earthquake induced liquefaction and deformation behaviour of an upstream tailings dam,” Advances in Materials Science and Engineering, vol. 2017, Article ID 5389308, pp. 1–12, 2017.View at: Publisher Site | Google Scholar
S. Mukerjee, D. Raj, and M. Bharathi, “Liquefaction susceptibility of tailings material,” in Proceedings of the 50th Indian Geotechnical Conference, Pune, India, December 2015.View at: Google Scholar
S. L. Kramer, Geotechnical earthquake engineering, Prentice-Hall, New Jersey, USA, 1996.
I. Towhata, “Geotechnical earthquake engineering. Series in geomechanics and geoengineering,” Springer-Verlag, Berlin, Germany, 2008.View at: Google Scholar
A. N. Schofield, “Dynamic and earthquake geotechnical centrifuge modelling,” in Proceedings of the Int. Conf. Recent Advances in Geotechnical Engineering and Soil Dynamics, UK, April 1981.View at: Google Scholar
Q. Chen, X. H. Chen, and X. Zhu, Tailings dam design handbook, Metallurgical industry press, Beijing, China, 2007.
Astm, Standard practice for classification of soils for engineering purposes (unified soil classification system), ASTM D2487, West Conshohocken, PA, 2011.
V. M. Zobin, Introduction to Volcanic Seismology, Elsevier, Amsterdam, Netherlands, 2012.View at: Publisher Site
D. Tanner and C. Brandes, Understanding Faults: Detecting, Dating, and Modeling, Elsevier, Amsterdam, Netherlands, 2020.View at: Publisher Site