Post-Evaluation of Slope-Cutting on Loess Slopes Under Long-Term Rainfall Based on Model Test

9 Failures of treated slope occurring in China are at a consistently increasing rate, leaving the huge number of treated loess 10 slopes calling for post-evaluation, however, no mature technique is in place. Depended on an loess slope in Shaanxi province 11 treated by slope-cutting, indoor geotechnical and model tests were conducted, revealing the rainwater infiltration 12 characteristics and pressure varying characteristics inside the slope, the results of which were then adopted to perform the 13 post-evaluation of the treated slope. The results showed that the rainwater scouring effect on the loess slope surface 14 attenuates gradually, and enters a steady stage after the first year of rainfall. The rainwater preferentially penetrates the 15 platforms with gradually attenuating rates, however the wetting front can not be deemed as the boundary between the 16 saturated and unsaturated areas, as the most parts of the model slope were indicated unsaturated by the pore water pressure 17 sensors. Caused by the in-situ stress release, the soil pressures don’t increase but decrease sharply at the start of the rainfall. 18 The displacements mainly occurs in the first two years of rainfall, following by steady periods. The model test results and 19 investigation results were then used to conduct the post-evaluation of the prototype slope, which formed a post-evaluation 20 frame relevant to other slope post-evaluations.


37
In recent years, more and more scholars devoted to the field of slope stability and failure mechanism investigation 38 under rainfall, incorporating modelling test 8-11 , numerical simulation 12-15 , and filed monitoring [16][17][18] . As revealed, slope 39 safety is critically influenced by rainfall, leading to the surface erosion to the slopes by rainwater runoff 19-23 , degrading the 40 slope soil strength as percolating into the slope 24-26 , and lessening the effective stress in the slope soil while saturating 41 it 27-31 . As the loess slope is loose, the rain drops and the runoff water can easily scour the soil particles of the slope surface, causing the lifting of the ground water table, which may eventually cause the collapse of the slopes 35 . Moreover, as a positive factor to the frictional strength of the loess, the effective stress favorable to the stability of the loess slope 49 decreases with the increase of the pore water pressure in the rainfall process 36 . Summarily, the rainfall is a critical factor 50 inducing the instability of slopes. Within the circumstances of engineering design, the water drainage measures should be 51 the obligatory practice to slope treating projects to minimize the hazards induced by rainwater percolating.

52
As the most convictive way to study the slope stability and failure mechanism, it was taken by many researchers in this 53 field. Schenato et al. 37 brought an optical fibre sensing system into slope model test to measure the pore water pressure, 54 the water content and the strain in the model slope during rainfall. The results revealed that the general evolution of the 55 slope during rainfall comprised of 4 stages which validated the effectiveness of the fibre sensing system in slope model 56 test. Lan et al. 38

68
Aiming at the repairing works of the treated loess slopes widely distributed in the north of Shaanxi, a loess model slope 69 shrunk by 10 times from the prototype slope treated by slope-cutting was built to study the treating effect. In the model 70 test, the rainwater infiltration process and the variations of pore water pressure, and soil pressure and displacement at the 71 key positions were recorded, which were in turn used to conducted the post-evaluation of the prototype slope. The formed 72 method of post-evaluation can be adopted in the treating effect assessments of other type of slope, while the rainwater 73 infiltrating characteristic and the variations of pressures and displacements in the slope have facilitating effect in the 74 further researching and designing works at this aspect.

75
The Prototype Failure

85
Located on the loess plateau, this project is a loess slope treated by slope-cutting with dwelling houses on the crest.

86
That is, the soils of the slope are wholly loess formed in late pleistocene epoch. As Fig. 2 shows, the height of the slope is 87 17.6 m, cut into three grades with the identical gradient of 56°. The height of the first grade is 5m, the height of the 88 second grade is 4m and the height of the third grade is 5.6m.
From field investigation, this project was built around the year 2012, having been run for about 5 years when the filed investigation. After a long period of operation, affecting by the rainfall, the slope was found collapsed in the right side 91 (see Fig. 2). To perform the post-evaluation of this project by model test, the simulated time must be 5 years according to   92   the actual project operation time frame. Depending on the similarity theory, the test time can be shortened by 100 times,   93   letting this model test be completed in a more tolerable period

99
Adopted from the research of Liu 42 , a 1 m wide, 2.5 m long and 1.8 m high model box was employed in the experiment 100 (see Fig. 3a). The left and right walls of the box were made from organic glass, letting the displacements in the slope and 101 the wetting front visible. The base and back walls of the model box were made from plank. And the the lower 30cm high 102 front was blocked by a plank which was drilled holes to vent the rainwater infiltrated into the slope. The upper plank of 103 the front could be removed to let slope surface unrestrained.

104
A frame made from steel pipes drilled holes of 1mm in diameter on one side was used as rainfall simulator. The valve 105 connected to the water pipe (see Fig. 3b) was used to controll the rainfall intensity. Prior to the start of rainfall, the rainfall 106 intensity could be calibrated to the certain values, which were achieved by a beaker and a measuring cylinder. in. When the model box was filled to the required height, the upper front plank was removed, then the model was cut to 126 the shape corresponding to the prototype.

127
As Fig. 5 illustrates, the model construction process was considerably complex, which can be delineated as follows:

128
(1) The procured loess was prepared with a precise water content as the prototype (ω=17.4%), and packed into the 129 model box with a certain quantity as a layer, then each layer was compacted to a thickness of 10 cm to control the density 130 of the slope model.

131
(2) In the model construction process, the pore-water pressure sensors and soil pressure sensors were buried in the 132 designed position.

133
(3) Synchronously, little colored sand particles were buried at the desired position close to the left and right walls of 134 the model box as the inner displacement marks.

135
(4) As the model box was rammed and filled to the desired height, it was then excavated to the shape corresponding

146
It is noteworthy that the soil properties were obtained from indoor tests according to Liu 42 . The water content of the 147 model soil was measured by the drying method: a certain mass of the loess was weighed before and after being dried in an 148 oven for 8 hours under 105~110℃. The water content was derived as the ratio of the loess mass lost to the loess mass 149 after dried. The permeability coefficient was tested by the standard variable head permeability test: firstly, installed the 150 loess specimen into the penetration container and sealed. Then, the the specimen was saturated in the container.

151
Thereafter, filled the head pipe with water to the proper water head. Opened the inlet of the penetration container, and recorded the the water heads in the head pipe and the corresponding time. The following equation was used to calculate the permeability coefficient.
Here, L is the specimen height, a is the section area of the head pipe, t1 and t2 are the starting and end timings of the water 156 head variation respectively, H1 and H2 are the starting and end water heads corresponding to t1 and t2. The density of the 157 soil sample was determined by the standard ring knife test: the prototype loess was sampled as ring specimens and 158 weighed by the scale. The density of the sample was the ratio of the specimen mass to the specimen volume in the ring 159 knife. The compression modulus of the loess was determined by the consolidation test: the standard specimens were 160 installed into the osmotic pressure container firstly. Then the different specimen heights at corresponding pressures were 161 recorded while the compressing settlement was stable. The varying void ratios derived from the gravity of soil particles, 162 initial density of the specimen, the initial water content and the specimen heights at different pressures were used to  Table 2.

Parameters Similarity relation Similarity constant Geometric dimension (L)
CL 10

178
It should be noted that the similarity constant of rainfall duration was not derived from π theorem, but from the

192
Here, tm and tp are the consolidation time of the model slope and the prototype respectively; Hm and Hp are the dimensions 195 of the model slope and the prototype respectively; Cv is the identical consolidation constant of the prototype and the 196 model slope. Thus, the similarity constant of rainfall duration can be calculated as : In the current study, CL equaled to 10, deriving Ct equaled to 100. That is, the test duration was shortened by 100 times,

210
To measure the displacements of the three slope shoulders, three key points were marked by nails inserted into the 211 corresponding positions, which are illustrated by Fig. 7 as S5, S4, S6 sequentially. S5 was located on the shoulder of the 212 first grade of slope, while S4 and S6 were located on the shoulders of the second and third grades of slope respectively.

213
To measure the displacements of the inner positions, three points in the slope close to the right wall of the model box was 214 marked by colored sand particles which are S1, S2, S3 at the identical depth of P1, P2, P3 respectively as Fig. 7 shows.

215
Meanwhile, three points in the slope close to the left wall of the model box at the same depths corresponding to S1, S2, S3 216 were marked as S1′, S2′, S3′ respectively in the same manner. The displacements of the key points were derived as the 217 differences of distances before and during rainfall, which were measured by a laser rangefinder seated a fixed position.

218
The accuracy of this laser rangefinder is up to 0.01mm which could satisfy measuring requirement without any doubt.

225
To capture the deformation process of the slope and the rainwater percolating process in the slope, a camera was used 226 to take photos from the front and the sides of the model slope at constant intervals.

239
Generally, the experimental process followed the steps below: 240 (1) The initial values of the soil and water pressure sensors were measured before they were buried in the constructing 241 of the model slope.

242
(2) When the built of the slope model was accomplished, the sensor cables were connected to the strainometer which 243 was connected to the computer to record the pressure data. Also, a laser rangefinder was fixed at a position in front of the 244 slope model and was connected to the computer to get the distances between the position and the displacement points (S1, 245 S1′, S2, S2′, S3, S3′, S4, S5, S6), thus to derive the horizontal displacements of the 6 points.

246
(3) The rainfall intensity of the simulator was calibrated to 31.94 mm/h, and acted on the model slope when the 247 computer program started.

248
(4) The steps in Table 3 was repeated 5 times to simulate the 5 years of operation while the computer program 249 recorded the soil pressures, pore water pressures and the horizontal displacements.

250
Results and discussion 251 Slope Scour Failure Process. By the camera fixed in front of the slope model, the slope scour failure process in the 252 rainfall was captured at certain timings. Depending on the field investigation, the simulating period was 5 years, 253 responding to the long-term rainfall effect on the slope. However, the scouring effect of rainfall after the third year was 254 negligible, thus is not presented in this section.

314
Rainwater Infiltration Characteristics. Rainwater infiltration can saturate slope soils, thus increasing the pore water 315 pressure, which significantly affects the stability of the slope. For slope-cutting treated slopes, the infiltrated rainwater 316 would reduce the internal friction angle of the slope soil, which is adverse to the treating project. Additionally, the 317 infiltrated rainwater causing the large water pressure in the slope body is also adverse to stability of the project. Some 318 researchers studied the influence of rainfall on the slope stability and found that the rainfall was the most significant 319 factor affecting the stability of slopes, the rainwater penetration dominates the deforming process of slopes 49,50 . In this 320 study, the two side walls of the model box were of transparent glass. Thus, the advancement of the wetting front could be 321 captured by cameras from the sides. 322 Fig. 11 presents the wetting front advancement with the rainfall process in the first year. It is clear that the rainwater 323 preferentially penetrated the platforms vertically, which could be due to the water accumulated upon the platforms. The 324 rainwater accumulated upon the platforms produced water heads there, actuating the rainwater into the soils vertically in a 325 larger rate. When the first rainfall lasted for 10 minutes, the penetration depths under the three platforms were identically 326 8 cm. When the first rainfall lasted for 20 minutes, the penetration depths under the first grade of platform, the second 327 grade of platform and the third grade of platform were 9 cm, 11 cm and 10 cm respectively. When the first rainfall lasted 328 for 40 minutes, the penetration depths under the first grade of platform, the second grade of platform and the third grade 329 of platform were 11.3 cm, 11.5 cm and 12 cm respectively. Thus we could derive that the vertical penetration rate of the 330 rainwater decreased with the proceeding of the rainfall. The average penetration rate was about 0.8 cm/min in the first 10 331 minutes of the rainfall, which decreased to about 0.2 cm/min when the first rainfall lasted for 20 minutes, then decreased 332 to about 0.075 cm/min when the first rainfall lasted for 40 minutes. It is reasonable that, the pore air in the shallow layer 333 of the slope soils is easier to be discharged by the infiltration of rainwater, thus giving more unblocked paths to the 334 penetration of the rainwater, which caused the higher penetration rate at the beginning of the rainfall. Eventually, when 335 the first rainfall lasted for 120 minutes, this rainfall was ended, showing an average vertical penetration rate of about 336 0.090 cm/min. Summarily, the wet front advancing rate decreased gradually in the first rainfall of the first year, eventually 337 stabilizing at the value of 0.090 cm/min. Similarly, we could derive the wetting front advancing rate after the rainfall. 30 minutes after the end of the first 339 rainfall, the average wetting front advancing rate under the platforms was approximately 0.07 cm/min. 5.2 hours after the 340 first rainfall, the vertical penetration depths under the first grade of platform, the second grade of platform and the third 341 grade of platform were 25 cm, 36 cm and 38 cm respectively, delivering the average wetting front advancing rate of about 342 0.026 cm/min. From the above, we can conclude that the wet front advancing rate after the rainfall is less than that during 343 the rainfall, with a decreasing trend over time. Further more, we could found from Fig. 11c that the penetration depths 344 under the slope shoulders were greater than that under the slope faces and platforms. It should be attributed to the dual 345 effects of penetration from the platform and the slope face, which could be deemed as the superposition of the two effects. Chueasamat 50 . For other key points, the most obvious is that the pore water pressure value of U1 decreased insistently 407 from -10.35 kPa to about -12 kPa, and the pore water pressure value of U4 declined insistently from -6.05 kPa to about 408 -12 kPa with a more drastic decreasing rate in the early years. As U1 was situated at the deepest location, the infiltrated rainwater was unable to recharge it while the soils under it absorbing the water nearby, causing the saturation degree of decreased in the rainfall process. As U4 was situated near the shoulder of the second grade of the slope, it was exposed in 413 the air after the first rainfall of the first year when the second grade of shoulder was ruined by the rainwater. That caused 414 the abnormal increasing of suction of U4 meaningless to this research. However, U2 and U3 situated at the mediate depth 415 of the model slope, the rainwater recharging effect from the above soils and the absorbing effect from the soils below kept 416 in balance, leading to no perceptive change of the pore water pressure variation there.

417
Moreover, from Fig. 13 we could find that the model slope was completely wetted by the rainwater, but the most part 418 of the model slope except U5 showed negative pore water pressures from Fig. 14, implying the unsaturated condition.

419
Thus, we can conclude that the wetting front can't be deemed as the boundary between the saturated and unsaturated areas 420 in rainfall processes. Accordingly, the developed Green-Ampt infiltration model 53 deemed that after rainfall the rainwater 421 continually migrates inducing a unsaturated area behind the wetting front. Thus, this developed Green-Ampt model is 422 supportive to the findings above.

454
The rainfall period of each year was comprised of 3 times of rainfall, with each time lasting for 2 hours.

455
Displacements of the Key Points. The distances between the 6 key points (S1, S2, S3, S4, S5, S6) and a fixed point 456 were measured by the laser rangefinder, the differences of which before and after the start of rainfall were the horizontal displacements for the corresponding points. Fig. 16 shows

477
In the second year, the displacement of S5 fluctuated up to about 14 mm and kept fluctuating around this value in the 478 following years. And the displacements of S2, S3 fluctuated around 2.5 mm and 9.2 mm respectively, while the 479 displacement of S1 still fluctuated around 0 mm. Thus, we could conclude that the point S1 kept stationary in all the five years of rainfall. For the points in a vertical line with S1, the larger upward distance from the point S1 in the slope model, 481 the larger displacement with it. That indicated the potential sliding surface went between S1 and S2.

487
Post-evaluation was brought into slope treating project in China by Zheng 4 , proposed the definition of post-evaluation of 488 slope treating. In pre-evaluation, engineers just concern about the safety of the project after the completion of construction.

489
Whereas, in post-evaluation they usually focus on the slope safety after a long term of operation and the running situation.

490
As Zheng addressed, the displacement rates of key points of the slope not greater than 0.1 mm/day could used as an 491 indicator of the slope safety. In addition, corresponding to the safety factor, a compound safety factor of slope was 492 adopted to judge the treating effect of the slope, which is detailed in      Geo-studio. Clearly, the sliding surface went between S1 and S2, which is strictly consistent with the model test results.

520
In the field of geotechnical engineering, the safety factor of slope is defined as the ratio of the total resisting moment to 521 the total sliding moment, as Eq. (7). 522 Figure 17. The critical sliding surface generated by Geo-studio. S1, S2, S3, S4 denote displacement markers identical with Fig. 7

543
To calculate the safety factor of the current slope, the sliding mass must be divided in to slices firstly. Simply, the 544 sliding mass of this project was divided into 6 slices, as illustrated in Fig. 18. To get the soil pressures and the suctions at 545 the bottoms of the slices, we assumed that the positions with the same burying depth have the identical soil pressure and 546 suction in this slope. Then, the soil pressures and suctions at the bottoms of slices could be derived by interpolating the 547 values of the adjacent depths. However, the soil pressures and pore water pressures at the toes of the three grades of slope 548 were different. Thus, the measured data of the points beneath the slope toes were used to derived the soil pressures or 549 suctions of the slices nearby, if it could be. By this way, the soil pressures and suctions at the bottoms of the six sliding 550 slices were derived as Table 5 shows.

558
It is clear that the safety factor derived from the proposed method combining with the measured data of model test was 559 largely greater than the critical value 1.2, indicating the conservative designing method of the loess slope treating project.

560
By inferring, it should be due to the underestimation of the soil strength in the slope designing method of China, which 561 didn't incorporate the matric suctions in the unsaturated condition. However, the derived safety factor here shows a 562 consistent result with that of the displacements of the model slope. Referring to Table 4, the treating effect of the 563 prototype project should be very good preliminarily.

564
Evaluating Results and Post-evaluation Frame. Synthesizing the above evaluating results, the slope was globally 565 stable having no further sliding trend, with relatively huge destruction caused by rainwater scouring. From that, the 566 collapse of the prototype slope was a local destruction. As a result, the treating effect of the project should be not bad, 567 reasonably.

568
Lastly, it is noteworthy that the post-evaluation frame formed in this paper are valuable to other slope treating projects 569 incorporating slope-cutting. Thus, this post-evaluation frame is detailed in Fig. 19. 570 571 572 Figure 19. Post-evaluation frame for slopes treated by slope-cutting.

575
To reveal the influence of long-term rainfall on the loess slopes in Yan'an city of Shaanxi province treated by 576 slope-cutting, filed investigations and indoor model tests were conducted, the results of which were employed to perform 577 the post-evaluation of the loess slope treated by slope-cutting. The following conclusions were obtained:

578
(1) The rainwater runoff induces serious scouring to the slope surface with the main patterns of gullies and shallow 579 sliding. For the critical reason of consolidation of the shallow layers in the intervals of rainfall, the scouring effect of 580 rainwater runoff becomes weaker and weaker with time, leading to the relative steady stage after the first year of rainfall.

581
(2) The rainwater preferentially penetrates the platforms with large rates for the rainwater accumulated there. During 582 the simulating duration, the wetting front advancing rate decreased gradually. 5.2 hours after the second rainfall of the 583 third year, the whole model slope was wetted with a wetting front advancing rate of about 0.003 cm/min.

584
(3) Though the whole model slope was wetted, the pore water pressure values of the prescribed points in the model test 585 were negative over the test duration, indicating the unsaturated condition. That is, the wetting front can not be deemed as 586 the boundary of the saturated and unsaturated area in the rainfall process. And the infiltrated rainwater preferentially 587 accumulates at the toe of the first grade of slope.

588
(4) Induced by the in-situ stress release, the soil pressures in the model slope drastically decreased in the first years of 589 rainfall, especially for the deeper points in the model slope.

590
(5) The horizontal displacements of the prescribed points in the model slope increased significantly in the first year of 591 rainfall, with a decreasing rate. After the second year of rainfall, the horizontal displacements had no regular increment, 592 implying the stable state of the slope.

593
(6) The above results were employed to perform the post-evaluation of the slope-cutting treating project of the loess