Similar to other research by model experiments, the variation in water content in soils during rainfall and earthquake tests in this study was depicted by saturation ratio Sr/Sr0, which is the saturation degree Sr normalized by its initial value Sr0. Along with that, the reduction in shear strength of the embankment was evaluated through the pore water pressure ratio Δu/σv′. In this, the change in pore water pressure Δu was normalized by the effective overburden pressure σv′ (calculated with the measured water content).
3.1. Effect of Fines Content on Slope Stability during Rainfall
First, rainfall-only experiments were conducted with a rainfall duration of 30 min for all three samples to find out the rainfall-induced failure time. The influence of fine particle content on the mechanical behavior of volcanic embankments during rainfall is discussed below.
Kawamura and Miura (2014) [
6] found that the increment of finer content in the Komaoka volcanic soil after rainfall experiments was nearly zero, implying that there is almost no particle breakage due to rainfall for this soil. However, because of their light weight and no cohesion, the fine particles easily move in the pores between the coarse particles under the effect of rainwater. In this study, an attempt to understand the trend of this migration was made by examining the grain size distribution with sieving tests in different parts of the embankments (Areas 1, 2, 3) and in the washed-out part (Area 4), as shown in
Figure 11. The results shown in
Table 6 imply that the fine particles in all three types of soil tend to move from Area 1 to Area 2 and 3 because of gravity and rainfall. It can also be seen that the fine particle content of Area 3 was greater than those of Area 2 in all three cases even though the direction of rainwater is towards Area 2 (shown by the movement direction of the wetting front in the next section). This may be due to fine particles moving with rainwater due to the piping phenomenon in the saturation region from Area 1 to Area 3. When the secondary failure had not yet occurred, the soil in Area 4, which was generated by runoff, consisted mainly of fine particles.
The variation in saturation degree during the rainfall tests is shown in
Figure 12. Under the effect of rainfall, the value of each meter started to increase at different moments depending on their locations. All three samples showed a similar trend: sm6, sm3, and sm1 located closest to the surface increased first; then sm5 and sm2 increased, and finally, sm4 increased. If the meters lie on the same wetted front, their values start to increase together and show the same trend, as noted in sm1, sm3, and sm6 or sm2 and sm5. At the time of failure corresponding to shear strain reached 6%, the saturation conditions inside the embankment in three cases were also similar: sm1, sm3, and sm6 increased to near steady values while the other meters remained at the original value. However, as the rainfall continued, the different hydraulic conductivity between the three soils caused the difference in the speed of the wetting front in the embankment. This difference can be more easily observed in the water content distribution shown in
Figure 13, which was obtained by linear interpolation from the six measurements inside the slope and the direct measurements at the boundary of the embankment before and after the experiments. In the figure, sample K
8.5A showed the highest permeability: after 30 min of rainfall, except for the deepest meter sm4 which increased to a rather high value, the meters sm1, sm2, sm3, sm5, and sm6 increased to their maximum value, as shown by the fact that they no longer increase when the rainfall stopped. For K
soil, sm4 did not start to increase after 30 min of rainfall. In particular, K
40A showed the lowest permeability: sm2 and sm5 did not even reach their maximum value. The effect of fine particle content on the hydraulic conductivity of volcanic soil in this study seems to be more obvious than it was in [
25], which can be explained as follows: Under the same dry density in [
25], the loss of fines content only leads to an increase in the average size of pores. However, a decrease in fine particle content directly results in the void ratio under the same compaction degree condition in this study. Therefore, the decrease in permeability in the latter case was more significant. After the rainfall stopped, sample K
8.5A showed the fastest drainage rate (sm6 did not decrease because it was exposed after failure), followed by K
soil and the slowest, which was K
40A.
The change in pore water pressure during rainfall is shown in
Figure 14. In the case of K
8.5A and K
soil, pw1 near the crown of the embankment increased faster than pw2 and pw3. Then, flow deformations occurred at the base of the slope when pw3 exceeded 1. For K
40A, the rate of increase in pore water pressure was lower than that of K
8.5A and K
soil. The values at three positions, pw1, pw2, and pw3, were almost the same. When the shear strain reached 6%, the pore water pressure was quite small compared with the other two cases. Another difference of K
40A was that the value of pw3 did not decrease for 3 h after the rainfall had stopped. The saturation degree sm4, sm5, and sm6 of K
40A also showed an upward trend after rainfall instead of decreasing as was the case of K
8.5A and K
soil. This can be explained through the movement of fine particles. As mentioned above, the fines content in Area 3 was always greater than in Areas 1 and 2. This migration trend led to the concentration of a large amount of fine grain at the toe of the slope. Fine particles combined with coarser particles made their size larger than the size of the pores, which prevented water and soil from continuing to move through that pore. This blocking effect in the basement made it difficult to drain and thus increased the groundwater table in the embankment as can be seen in the K
40A case.
The increment of shear strain during rainfall is shown in
Figure 15. When the shear strain was less than 6%, these relationships were almost linear for all three samples. Failure time for K
8.5A and K
soil was 10 min and for K
40A was 9 min. Since the increase in pore water pressure during this period of K
40A was smaller, the faster failure rate of this soil sample is considered to be due to its smaller shear strength (internal friction angle). Thus, although an increase in fine content was shown to reduce soil permeability, it had almost no effect on rain-induced slope failure when the shear strain is less than 6%. After the shear strain exceeded 6%, it sharply increased at 27 min for the K
8.5A embankment along with secondary failure, which was not observed for K
soil and K
40A during 30 min of rainfall. Even in another experiment when the precipitation time was increased to more than 1 h, secondary failure did not occur with K
40A. This difference can be explained through the failure pattern of K
8.5A and K
40A shown in
Figure 16. In the case of K
8.5A, along with the increment of pore water pressure, the flow deformation that occurred at the base of the embankment led to the fast development of a larger slip line. However, the other parts in the model slope except for the slip line showed almost no change. In contrast, the K
40A embankment did not show slip line or gully erosion, but a small local failure appeared on the entire slope surface leading to a gradual reduction in size. Even when the pore water pressure exceeded 1, the flow deformation did not occur. The surface failure for K
40A can be attributed to the fine grains migration in the embankment as described above, because fine particles play an important role in the construction of K
40A’s soil structure, not just filling the voids as in the case of K
8.5A. When the soil was gradually saturated, the flow of water inside the embankment became more difficult, leading to a decrease in the movement of fine particles. As a result, the rate of increase in shear strain for K
40A was reduced.
3.2. The Failure of Embankments with Different Fine Contents under Post-Rainfall Earthquake
In
Figure 17, the input motion of the 2.8 m/s
2 seismic loading used in the earthquake-only experiments [
6] is shown through the measured acceleration at the shaking table (point B), the upper part (point A), and the lower part (point C) of the embankment. It can be seen that the maximum acceleration near the crown area of the slope was larger than those at other locations. As the fine content varied, the samples in earthquake-only tests of this study showed similar mechanical behavior and shear strain (less than 2% after 400 cycles), implying that fine particle content had no significant effect on the seismic resistance of compacted volcanic soils at a compaction degree of 90%. However, this behavior is no longer correct in the post-rainfall earthquake experiments presented below.
The cumulative rainfall up to the time of failure mentioned in
Section 3.1 was set as the total rainfall amount R
0:
- -
For K40A: R0 = 100 mm/h × 9 min = 100 mm/h × 0.15 h = 15 mm;
- -
For K8.5A, Ksoil: R0 = 100 mm/h × 10 min = 100 mm/h × 0.17 h = 17 mm.
In this section, the model embankment was first subjected to the precipitation of R equal to 0.5R0. As a result, the rainfall-induced shear strain was 3.88% for K8.5A, 4.69% for Ksoil, and 3.11% for K40A. After the rainfall had stopped for 90 s, seismic loadings were applied when the residual pore water pressure was still high. The total number of cycles was 100 divided into five applying times, each time was about 75 s apart.
The response acceleration due to post-rainfall earthquakes at different locations inside the model embankment is shown in
Figure 18. The previous rainfall caused this behavior not to retain the periodicity as in the earthquake-only experiment, but the acceleration at the upper parts was always higher than those at the lower parts of the slope. The maximum value of the acceleration at the crown area in the case of K
8.5A, K
soil, and K
40A were 5.51, 4.43, and 4.33 (m/s
2), respectively. It can be seen that these values were higher than the value of 3.2 m/s
2 in the case of the earthquake-only experiment. This is considered to be due to the deformation and the change in soil structure generated by previous rainfall.
The behavior of saturation degree under the post-rainfall earthquake is shown in
Figure 19. As can be seen, the variation during rainfall was similar to that of the beginning of the rainfall-only experiment: sm1, sm3, and sm6 started to increase due to the infiltration of rainwater but did not yet reach their maximum value while sm2, sm4, and sm5 were almost unchanged. When cyclic loadings were applied, a general trend was observed for all three samples: sm1, sm3, and sm6 increased while sm2 and sm5 decreased. In which, the change in sm1 and sm2 was more obvious than that of sm6 and sm5. Under the effect of seismic loadings, the particles rearrange, and the pore size changes, leading to the movement of water in the pore. Through the change in the distribution of water content before and after the earthquake as shown in
Figure 20, we can see that the cyclic loadings have the effect of pushing water from the inside to the outside of the model embankment. Another point worth noting for the K
40A case was the difficulty of drainage after earthquakes, which occurred not only in the basement but also in the upper parts of the slope. Thus, cyclic loadings can cause a blocking effect due to fine particle concentration at any point in the K
40A embankment.
The change in pore water pressure during post-rainfall earthquakes is shown in
Figure 21. Similar to the rainfall-only experiment, the measured values at the positions of pw1, pw2, and pw3 both increased during rainfall, and the difference between them was small. During earthquakes, the pore water pressure changed periodically. As the shaking table begins to move, it forces the slope back causing volume contraction and a decrease in voids, increasing pore water pressure. When the shaking table returns to the original position, the soil on the slope has a large degree of freedom, the pores between the particles increase, so the pore water pressure decreases. The amplitude of this periodical change in pore water pressure was greatest in the case of K
8.5A.
Figure 21 ignored these very rapid variations and focused on the changing trend of pore water pressure after earthquakes. As presented in the above section, earthquakes cause pore water to move outwards. Therefore, pw1, pw2, and pw3 placed near the slope surface increased together under the effect of cyclic loadings. It can be seen that the increase in pw1 near the crown of the slope was larger than that of pw2 and pw3. In the case of K
8.5A, the pore water pressure at pw1 exceeded one at the first time of seismic loading and the flow deformation near the crown caused slope failure. During the 4th seismic loading, the pore water pressure of pw3 increased suddenly, causing the shear strength to rapidly decrease, the two soil sections were separated, creating cracks that were recorded near the basement. Pw1 in the case of K
soil reached a value of one at about the 70th cycle, corresponding to the total shear strain due to rainfall and earthquake surpassing 6%. In the case of K
40A, pw1 was still less than one after 100 cycles and only exceeded one at about the 140th cycle. This is because the shear strain due to rainfall was smaller in this case. As can be seen in most tests, the point of shear strain due to rainfall and earthquake reaching 6% was relatively close to the time when the pore water pressure ratio reaches one. Thus, the measurement of pore water pressure still plays an important role in assessing slope stability.
The increase in earthquake-induced shear strain of embankment subjected to previous rainfall is shown in
Figure 22. In the figure, K
8.5A was the fastest, and its rate was much higher than that of the other two samples. Meanwhile, the behavior of K
soil and K
40A was almost similar. The failure form of these two soils was also the settlement of the entire embankment due to the rearrangement effect of seismic loadings while the failure due to flow deformations and cracks were recorded in the case of K
8.5A as mentioned above. Soils with small fine particle contents have higher hydraulic conductivity and pore water pressure is easier to increase during post-rainfall earthquakes. Meanwhile, high fines content soils, which have a small void ratio, increased their density even more during rainfall. Therefore, seismic loadings have difficulty in rearranging the particles to reduce pore size and increase pore water pressure. In addition, the increase in density also makes the stabilizing force due to self-weight more advantageous than the earthquake-induced destabilizing horizontal force. However, the characteristics that increase this resistance to seismic loadings in K
soil and K
40A are similar. As the fine content increases and exceeds a certain limit, its effect diminishes because fine particles occupy most of the voids in the soil. For these reasons, the earthquake resistance of compacted volcanic soil is subjected to previous rainfall increases with the increase in fine particle content. However, this strength does not increase further after the fine particle content exceeds a certain threshold of about 27% in this study.