Liquefaction Characteristics of Sandy Soil Distributed in Wind Power Farms, Soc Trang Province, Vietnam

Abstract


Introduction
Wind energy is safe and renewable energy which has been widely utilized for the economic growth demand in many countries (Zheng et al., 2016).One of the biggest challenges when building a wind turbine is the impact of dynamic and cyclic loads on the foundation.For onshore wind turbines, the cyclic and dynamic loads acting on the foundations generate from wind, 1P (rotor frequency), and 2P/3P (blade passing frequency) (Nikitas et al., 2017).The soil properties such as long-term strength, and liquefaction potential can be changed under these loads and affects the long-term stability of wind turbine.In general, the cyclic strength and soil liquefaction potential can be indirectly evaluated based on the results of in-situ tests such as cone penetration test (SPT) (Seed and Idriss, 1971;Seed and Peacock, 1971;Iwasaki et al., 1982), cone penetration test (CPT) (Robertson and Wride, 1998;Ateş et al., 2014;Du et al., 2019), shear wave velocity (Vs) (Hatanaka et al., 1997;Tokimatsu et al., 1986;Andrus and Stokoe, 2000;Ateş et al., 2014) or directly assessed based on laboratory tests such as cyclic triaxial test, cyclic direct simple shear test, cyclic torsional shear test, and shake table test (Satyam, 2012).Regarding the design of wind turbine foundations, the soil properties under dynamic and cyclic loads have been widely investigated (Safinus et al., 2011;Sim et al., 2013;Blaker and Andersen, 2019).In general, these studies have indicated the important role of laboratory testing to determine the cyclic properties of soil for designing turbine foundations.
In Vietnam, many wind powers farms have been recently built and will be built in near future, especially in some provinces such as Soc Trang, Bac Lieu, Ninh Thuan, Binh Thuan, Ca Mau, Vung Tau, and the Central Highlands.According to the survey, Soc Trang's coastal areas have a great opportunity for wind power development since there is a plentiful and stable wind blowing every month of the year with the average wind speed at the height of 60 m between 6.2 to 6.4 m per second.According to the Ministry of Industry and Trade of Vietnam, 35740 hectares of land area in Soc Trang province has been investigated for wind power development with a potential capacity of 1470MW.The wind farm will be mainly distributed in three areas along the coastline, including Vinh Chau town, Tran De, and Cu Lao Dung districts.The planning areas of wind farms in Soc Trang province are mainly located on alluvial ground aged Holocene and Pleistocene.In which, the alluvial and marine sandy soils are common and distributed at different depths (Nu, 2014;Nu et al., 2020).These soils are sensitive to cyclic and dynamic loads, especially sandy soils.In which, the potential of soil liquefaction and reduction of soil shear strength under these loads is high and affects the stability of the foundation.Thus, investigation of the liquefaction potential of sandy soils in Soc Trang's coastal area is very important for ensuring the stability of the wind turbine foundation.
In Vietnam, the research on cyclic strength and soil liquefaction potential is still limited.Some of the previous studies regarding these issues are mostly focused on soils distributed in big cities such as Ha Noi and Ho Chi Minh cities (Binh et al., 2016;Phong, 2016;Phong andThang, 2014, 2016).Recently, Nu et al. (2021) investigated the liquefaction potential of sandy soils in the North Central Coast of Vietnam based on SPT values.In general, these studies are mainly evaluated based on earthquake-induced liquefaction.However, the dynamic load generated from wind turbine operation is different from earthquakes.Besides, there has been no research on cyclic strength and the potential for liquefaction of sandy soil distributed in wind power farms in Vietnam.In this study, the sandy soil samples taken from boreholes in Vinh Chau town, Soc Trang province will be used to assess the liquefaction potential under dynamic loads generated from wind turbine operation.The undrained cyclic triaxial apparatus will be employed to investigate the liquefaction potential of sandy soil.For sandy soil, it is difficult to take the undisturbed soil samples, so the remolded samples which simulate the field density of sandy soils were prepared for the cyclic triaxial test.

Sample Preparation
Sandy soil samples were taken from two boreholes (S1, S2) in two wind farms in Vinh Chau town, Soc Trang province.Borehole S1 (coordinates: X=615930; Y=1034064) is located in Lac Hoa wind power farm whereas S2 (coordinates: X=617345; Y=1040603) is located in Hoa Dong wind power farm.The location of two boreholes for soil sampling is shown in Fig. 1.According to the documents of geotechnical investigation, there are two types of sand distributed in the study area to the depth of 55m (S1) and 60m (S2).The characteristics of sand layers in two boreholes are listed in Table 1.As shown in this table, the sand layers are distributed alternating with clayey soft soil layers.In which, the very fine sand layer (layer 2) with a loose state is distributed from 2.5 to 13.0 m while the fine sand with a dense (layer 5) to a very dense state (layer 6) is distributed from 31.0 to 60.0 m.The average thickness of layer 2 is 3.5m whereas the average thickness of layers 5 and 6 is 8.0m and 10.0m, respectively.The standard penetration resistance (NSPT) is from 3 to 5 blows for layer 2; from 27 to 42 blows for layer 5, and above 50 blows for layer 6.Based on the field description of sand samples, the SPT values, and the results of particle analysis, it is seen that the sand layers in two boreholes (S1, S2) are almost identical.The underground water levels recorded in two boreholes varied from 1.0 to 1.6 m.Sand samples will be taken from 3 sand layers (Nos.2, 5, 6) for this investigation.Since the sand layers in two boreholes are almost similar, sand samples in each layer of two boreholes were well mixed for sample preparation.The sand mixture samples are denoted as C1, C2, and C3.Some physical properties of these sand samples are presented in Table 2.As shown, the maximum void ratios of tested sand samples are from 1.021 to 1.050 while the minimum void ratios range from 0.577 to 0.593.For sandy soil, the specimens for laboratory testing can be prepared according to four methods: dry and wet pluviations, and dry and moist tamping (Mulilis et al., 1976;Juneja andRaghunandan, 2008, 2010).The dry and wet pluviations, and dry tamping methods often produce specimens with a low density (high void ratio) (Juneja and Raghunandan, 2010).In this study, sand sample C1 was taken from the loose sand layer (layer 2), so the dry pluviation method will be employed for sample preparation.The sand samples C2, and C3 were taken from dense to very dense sand layers (layers 5 and 6), so the moist tamping method may be suitable for sample preparation of these sand samples.In the dry pluviation method, the dry sand is poured into the split mold with an inner diameter of 70 mm and a height of 140 mm (Fig. 2) using a pluviator device.The density of the sample is controlled by the amount of dry sand pouring into the mold.In the moist tamping method, after determining the amount of dry sand based on the desired density, it will be mixed with de-aired water at a moist content of 2%.To ensure a uniform density of the sample, the amount of dry sand will be divided into seven equal parts.Each part is well mixed with de-aired water to the moist content of 2% and then placed into a split mold by the spoon and carefully compacted at a fixed height of 20 mm (the total height of 140 mm).The remold sample after preparation is shown in Fig. 2. In this study, sand samples will be remolded using the methods mentioned above to simulate different distribution depths, especially for sand in shallow depths such as layer 2. Some properties of sand specimens before testing are listed in Table 3.As presented in this table, initial relative density (Dr) of samples C1 represented for layer 2 (loose state) ranges from 0.180 to 0.287.For sand samples C2, and C3 (represented for layers 5 and 6), the values of Dr range from 0.643 to 0.790.

Testing Procedure
The cyclic triaxial test apparatus (Tritech 50 KN -Controls Group) is employed for this investigation.The testing procedure was conducted in accordance with ASTM D5311 (ASTM-D5311, 2013) and included three main stages: saturation, consolidation, and cyclic loading.For the saturation stage, the back-pressure saturation method is used to saturate the sand specimens in this study.The difference in cell and back pressures is maintained at 10 kPa.The saturation of the specimen is confirmed based on Skempton's B-parameter.It is the ratio of the change in pore water pressure Du to a change in cell pressure Dsc in an undrained condition.The saturation process is completed when the B value is higher or equal to 0.95.Normally, the full saturation of very fine sand was obtained with a back pressure of 70 kPa.
For the consolidation stage, after the saturation condition of the specimen has been obtained, the specimen was consolidated under the effective confining pressure at which it would be tested.The consolidation process was conducted by increasing the cell pressure to the desired value and then opening the valve for drainage.The cell pressure was increased to reach the desired effective confining pressure which is the difference between the cell and back pressures.To simulate the field condition, the effective confining pressure will be set equal to the effective overburden pressure which is based on the simulated depth of samples and water table level.The parameters for saturation and consolidation stages are shown in Table 4.For the cyclic loading stage, the testing parameters were chosen based on types of construction works and dynamic loads.For the onshore wind power farms, the load caused by wind and operation of the turbine is cyclic and acts on the foundation for a long time, so the number of cycles required is high.The frequency (f) is often chosen in the range of 0.4 to 1 Hz (Bhattacharya, 2019).In this study, the frequency of 1 Hz will be applied throughout the tests.The stress amplitude (Dsa) was determined based on the desired cyclic stress ratio (CSR) as the following formula: Dsa= CSR ´ 2s¢c (s¢c: effective isotropic consolidation stress).The CSR can be determined based on two conditions: 1) Utilizing the maximum bearing capacity of the soil (CSRult), in this case, the CSRult can be estimated based on Nspt values; 2) According to the maximum stress value that the construction load is likely to cause in the ground.Under this condition, the greater the depth, the smaller the CSR values.

Results and Discussions
After completing the consolidation stage, the cyclic loading for all specimens was conducted with the same frequency of 1 Hz with different CSR values.The specimens at shallow depths and close to the dynamic source will have a high value of CSR and vice versa.The test results of 9 sand specimens are summarized in Table 5.
As shown in Table 5, the very fined sand specimens (C1-0, C1-1, C1-2, C1-3, and C1-4) are distributed from 5.5 to 11.5m having a relative density of lower or equal to 0.3 (loose state) whereas the fine sand specimens (C2-0, C2-1, C2-2, C3) are distributed at the depth of higher than 30m having the relative density greater than 0.7 (dense state).It can be seen that these results are consistent with the field conditions of these sand specimens.The results of the cyclic triaxial test on some sand specimens C1-2 (liquefied) and C2-1 (non-liquefied) are shown in Figs. 3 and 4, respectively.As shown in Fig. 4a, the pore pressure ratio (Ru) significantly increases in the first 15 cycles (increasing in density), after that it will increase to about 70% at the cycle of 225.Above 70%, Ru quickly increases and reaches 100% at the cycle of 228.The changes in stress and strain are insignificant with Ru of less than 70%.When Ru increases above 70%, the stress decreases while the strain increases significantly and exceeds 5% (Fig. 5a).This sand specimen is liquefied at the cycle of 228.The cyclic behavior of other sand specimens at layer 2 (C1-0, C1-1, C1-3, C1-4) is also similar to that of the sand specimen C1-2.For sand specimen C2-1, as shown in Fig. 4b, the values of Ru increase slowly with the increasing of cycles and reaches the value of about 18.6% at the cycle of 500.The strain as presented in Fig. 5b, significantly increases at about the first 100 cycles, then it slowly increases to about 0.5% at the cycle of 500.This sand sample is non-liquefied and its behavior is similar to that of sand specimens C2-0, C2-3, and C3.In the case of liquefied failure, as shown in Fig. 3a, when Ru increases to higher than 60-70%, there is a significant change in stress and strain.However, the liquefied cycles, the level of strain, and the increase of Ru in different specimens are different since the values of CSR applied for these specimens are different.Fig. 5 presents the relationship between Ru and the value of CSR.It can be seen that the changes of Ru during cyclic loading significantly depend on the CSR.The higher CSR value, the faster Ru increases, especially at CSR above 0.2 and the specimen is prone to liquefy (specimen liquefies at a smaller number of cycles).As presented in Table 4, when the CSR values increase from 0.16 to 0.234, the liquefied cycles decrease from 468 to 4 cycles.As shown in Tables 4 and 5, in the study area, the sand distributed at the depth of less than or equal to 11.5 m will be liquefied under the testing conditions.This result is consistent with that of previous studies.Satyam (2012) reported that the sand distributed deeper than about 15m is rarely found to be liquefied.The changes of Ru with the increase of cycles for different sand specimens are plotted in Fig. 6.As shown, there is a significant difference in Ru tendency between very fine sand specimens (C1-0, C1-1, C1-2, C1-3, C1-4) and fine sand specimens (C2-0, C2-1, C2-3, C3).For fine sand specimens distributed at a high depth, the CSR values are often less than or equal to 0.1.When the CSR is less than or equal to 0.1, the maximum increase of Ru is about 18.6% and the maximum strain is about 0.87%.For these sands, when the CSR is 0.23, the increase in Ru and strain is not significant.Thus, the fine sand at the depth of above 30 m seems to not liquefy under the testing conditions.However, the increase of Ru by 18.6% in fine sand can lead to a decrease in soil strength and affects the stability of the foundation when this sand is used as the bearing layer, especially for the pile foundations.Wada (2016) reported that in fine-grained soil, the excess pore water pressure generated at the pile toe during driving work could decrease the bearing capacity of the pile foundation.Therefore, the generation of excess pore water pressure in fine-grained soil under dynamic loads should be considered when estimating the bearing capacity of pile foundations.

Conclusions
Based on the analysis of cyclic triaxial test results, some conclusions are drawn as follows: • In the study area, the very fine sand distributed at less than or equal to 11.5m will be liquefied under the testing conditions.The liquefied cycles depend on the CSR values; it decreases from 468 cycles to 4 cycles when the CSR values increase from 0.16 to 0.234.Additionally, when the Ru increases to above 70%, the sand specimens will be quickly liquefied.• For the dense fine sand above 30.0m, it will not be liquefied under the testing conditions.However, the increase of Ru by 18.6% can decrease the soil strength and affects the bearing capacity of the foundation, especially for the pile foundation.Thus, the increase of pore water pressure under the dynamic loading should be considered when estimating the bearing capacity of the pile foundation.