In general, researchers agree that aeolian sands consist of fine to medium grained loose sandy soil. They are poorly graded sands, with great uniformity, which can be attributed to the selective sorting of the wind (Elipe and López-Querol 2014). The aeolian sand used in this study was collected from a transmission line construction site at Dalate Banner in the Hobq Desert in the Inner Mongolia Autonomous Region, which is a typical in situ soil in Northwest China.

The particle size distribution of the aeolian sand in this study was obtained according to ASTM D422 (ASTM 2007), as shown in Figure 2. The parameters that typically describe the shape of the particle size distribution curve—D ₁₀ = 0.09 mm, D ₃₀ = 0.12 mm, D ₅₀ = 0.15 mm, and D ₆₀ = 0.17 mm—indicate that the UAS in this study can be classified as SP, i.e., poorly graded sand, according to Unified Soil Classification System ASTM D2487 (ASTM 2011a). This conclusion is consistent with that reported by Elipe and López-Querol (Elipe and López-Querol 2014).

The average specific gravity of the samples was 2.67, which is similar to the mean specific gravity values for dune sands that have been determined in many other parts of the world. For example, specific gravity values for the aeolian sand in Libya are between 2.59 and 2.70 (Khan 1982), whereas those for aeolian sand range from 2.63 to 2.67 in Qatar (Al-Ansary et al., 2012), from 2.62 to 2.75 in Kuwait (Al-Sanad et al., 1993), and from 2.67 to 2.70 in Iraq (Al-Taie et al., 2013), respectively.

In most of the cases, the natural moisture content is extremely low, ranging between 0 and 4%, due to the infrequent rainfall, deep water table and high capacity of evaporation (Elipe and López-Querol 2014). As a result, in this study, the aforementioned aeolian sands were moulded to a moisture content of 3% for both UAS and CAS backfills.

The cement used in this study was ordinary Portland cement (OPC), which is widely used in China. In general, the cement requirement for low plasticity soils and sands varies from 3 to 11% by dry weight (Al-Aghbari and Dutta 2005). To determine an optimal cement content for CAS backfills, direct shear tests were carried out on specimens obtained from the CAS with cement content of 3, 6, and 9%, by dry weight of the sand. The aeolian sand with a moisture content of 3% and the required amount of OPC corresponding to different percentage of cement were mixed thoroughly to a uniform colour, respectively. The resulting mixtures were then compacted in the compaction mould and cured at room temperature. Standard loading and measuring procedures were conducted in accordance with the ASTM D3080 (ASTM 2011b) after 28 days curing time for CAS.

Each of the shearing tests was performed by preparing the specimens (20 mm in height and 61.8 mm in diameter) for the direct shear tests, applying a predetermined normal stress, pushing forward the frame of the shear box horizontally at a constant displacement rate of 0.02 mm/min, and measuring the shearing force and horizontal displacement as the specimen was sheared. Figures 3–6 show the nominal shear stress and normal displacement versus shear displacement of the specimens in direct shear tests for UAS and CAS with different cement content under different normal stresses (100 kPa, 200 kPa, 300 kPa, and 400 kPa), respectively.

The behaviour of the UAS and CAS backfills shown in these figures can be described as follows, with attention to the influences of normal stress:

For UAS with a moisture content of 3% and a relative density D _r = 60%, all curves of the nominal shear stress versus shear displacement in Figure 3 indicate gradually hardening behaviour (showing no peak shear stress), and all normal displacements increase with shear displacement indicating shear-compression (negative normal displacement). Thus, as the normal stress increases, the shear compression will also increase. Physically, it is likely that the interaction between the sand grains is linked to the rounded surfaces, which leads to a easier slip.

For CAS with cement content of 3%, the curves of the shear stress versus shear displacement in Figure 4 show hardening-softening (showing a peak shear stress) behaviour under normal stresses of 100 and 400 kPa, but indicate gradually hardening behaviour under normal stresses of 200 kPa, 300kPa. The corresponding curves of normal displacement and shear displacement show shear-dilation (positive normal displacement) under normal stresses of 100 kPa, 200 kPa, and 300 kPa and shear-compression for higher normal stresses of 400 kPa.

For CAS with cement content of 6% as shown in Figure 5, the curves of the shear stress versus shear displacement change from hardening-softening under a normal stress of 100 kPa to gradually hardening behaviour under normal stresses of 200 kPa, 300 kPa, and 400 kPa. The corresponding curves of normal displacement and shear displacement show shear-dilation under a normal stress of 100 kPa and shear-compression for higher normal stresses of 200 kPa, 300 kPa, and 400 kPa.

For CAS with cement content of 9% as shown in Figure 6, all curves of the nominal shear stress versus shear displacement exhibit gradually hardening behaviour. All normal displacements increase with shear displacement indicating shear-compression, and as the normal stress increases, the shear dilation increases.

In order to compare the effects of amount of cement on mechanical behaviour of CAS with cement content, shear strength parameters such as cohesion and angle of internal friction were evaluated from the tests. Using the experimental test data, the shear stress at failure was plotted against the normal stress, and a straight-line fit was used to determine the fitting parameters of angle of internal friction and cohesion. Figure 7 shows variation of shear strength parameters of angle of internal friction and cohesion with cement content, which provides a clear indication of the improved aeolian sand shear strength. However, it should be noted that the failure criterion was considered as the point at which the shear stress starts to decrease (peak shear stress) or to remain fairly constant after an amount of 4 mm in relative lateral displacement (critical shear stress), as observed in the raw test data.

As shown in Figure 7, the angle of internal friction for the UAS specimens was 36.1° and the cohesion was approximately zero. However, the corresponding values of the angle of internal friction and the cohesion were 37.8°, 30.2°, and 37.9°, and 18.2, 43.6, and 16.8 kPa for the CAS specimens with the cement content of 3, 6, and 9%, respectively. In general, the angles of internal friction of the UAS and CAS specimens had less dependency on the addition of cement. However, the maximum cohesion is 43.6 kPa when 6% cement was added to the aeolian sand with a moisture content of 3%. Therefore, the optimal cement content of 6% could be determined for CAS backfills in the next uplift load tests.

All the laboratory-testing program were carried out at the apparatus in Figure 1 described previously. The UAS backfill was also moulded to a moisture content of 3% by aforementioned aeolian sand obtained in the Hobq Desert in the Inner Mongolia Autonomous Region in advance, and the CAS backfill was prepared in a rotating drum mixer by mixing the aeolian sand with a moisture content of 3% and the required amount of OPC corresponding to the optimal cement content of 6% by weight of dry aeolian sand to a uniform colour.

Appropriate trench backfill and compaction is a critical success factor for load test. Controlled pouring and tamping techniques were used in tests to obtain homogeneous UAS and CAS beds of reproducible backfill. The unit weights of the sand bed in the tests were controlled by weighing the required quantities of UAS and CAS for each layer beforehand to ensure compatibility between the backfill volume and the weight placed in the compaction pit. Bulk density tests indicated that the average unit weights were 14.5 kN/m³ and 14.7 kN/m³ for UAS and CAS backfills, respectively. Overall, the backfill control indicated a homogenous sand bed for UAS and CAS, and each of the model plate anchors was embedded in these prepared UAS and CAS backfills.

Each of the load tests on the model plate anchors embedded in CAS was conducted after 28 days of curing time, whereas the tests on the model plate anchors in UAS did not have any curing period. All the tests were conducted with the same loading, reaction, instrumentation, and data acquisition systems.

The slowly maintained load method was adopted in all tests; that is, the uplift loading was applied in increments of 10% of the predicted ultimate load of each individual foundation, and the foundation was allowed to move under each maintained-load increment until a certain rate of displacement was achieved. Each load increment was maintained after loading until two consecutive displacements within each hour were less than 0.1 mm. Then, the next increment of load was added. This is the typical test procedure recommended by the CNS GB50007 (CNS 2011) and CLC JGJ 94 (CLC 2008). Each of the load tests was continued to the point at which failure was believed to have happened, and then, the test was halted.

In this study, the uplift load test results of all the tests in UAS and CAS backfills were discussed in terms of load–displacement curves. Figure 8 illustrates the plots that relate the applied uplift load to the pier-head vertical displacement for each of the model plate anchors.

In general, the load–displacement curves of the model plate anchors embedded in CAS differ markedly from those in UAS. As shown in Figures 8A–C, the load–displacement curves of the model plate anchors embedded in UAS generally exhibit a three-phase behaviour: 1) initial linear behaviour, which shows a gradual increase in uplift resistance; 2) nonlinear transition to peak uplift resistance behaviour; and 3) post-peak behaviour, which shows a slight, continuous decrease in resistance with larger displacement before reaching the residual phase. However, as shown in Figures 8D–F, the load–displacement curves of the model plate anchors embedded in CAS present a typical two-phase behaviour: 1) pre-peak behaviour, in which the uplift resistance increases linearly but shows only slight increases in displacement and the load-carrying capacity of the model plate anchor reaches a peak load; and 2) post-peak behaviour beyond the peak where the uplift resistance either decreases rapidly or increases slightly as the displacement increases.

It should be noted that the failure can be easily identified in all the tests on the model plate anchors embedded in UAS and CAS; the former started at large displacements and the latter appeared at tiny displacements. The uplift failure resistance at the peak point, T _UR, and the corresponding displacement, s _UR, for each of the load tests is summarised in Table 1.

Based on the comparisons in Figure 8, it is also clear that the initial pre-peak uplift stiffness benefits strongly from the addition of cement. This obviously highlights the effectiveness of the insertion of cement in aeolian sand for the improvement of the uplift performance of plate anchors. The addition of cement does, however, lead to a brittle load–displacement response. This phenomenon is consistent with that reported by Rattley et al. (Rattley et al., 2008). As a result, uplift resistance reduces sharply or displacement increases dramatically after the peak resistance is mobilised at an average displacement of 1.20% of the plate width, whereas the average displacement for the model plate anchors embedded in UAS is about 1.64% of the plate width. This may be related to the destruction of the cement bonds between aeolian sand grains with continued shearing (i.e., reduction in the cohesion component of shear strength). However, such brittleness may not be a significant design consideration for transmission tower foundations, which often limit uplift to 25 mm under ultimate loads; this uplift equates to a displacement of about 1% for typical foundation width of 2.5 m (Rattley et al., 2008).

Failure of the uplift-loaded plate anchors in UAS and CAS was typically associated with the appearance of cracks on the ground surface in the vicinity of the anchors when tensile strength is reached before failure (i.e., before reaching ultimate shear strength), as shown in Figure 9. Similar to the report by Consoli et al. (Consoli et al., 2013), it can be drawn that the failure surfaces were approximately straight lines that began close to the outer edge of the anchor plates and symmetrically propagated to the ground surface with a slope of the failure surface. In general, the failure of the plate anchors in CAS had a larger truncated cone shape with a slope angle with the vertical and wider brittle radial cracks than those of the plate anchors in UAS.

Thefailure of vertically uplifted foundations can be classified into shallow and deep modes (Meyerhof and Adams 1968; Pacheco et al., 2008). According to the studies by Pacheco et al. (Pacheco et al., 2008), the model plate anchors embedded in UAS and CAS in this study would be in shallow failure mode. As a popular method, the earth cone or dead weight method has been widely adopted for plate anchor designs worldwide (Matsuo 1967; Meyerhof and Adams 1968; Matsuo 1968), and Figure 10 shows the assumed slip surface of this method in the shallow failure mode. In this method, the ultimate uplift resistance is assumed to be equal to the sum of the dead weight of the foundation and the soil mass contained in the truncated pyramid with the bottom of the foundation slab as a base; therefore, the theoretical equation for the uplift resistance can be determined using the following relationship: T = γ s ( V t - V 0 ) + G f (1) where G _f is the dead weight of the foundation, γ _s is the unit weight of the earth lying above the bottom of foundation slab, V ₀ is the volume of the foundation below the ground surface, and V _t is the volume of the truncated pyramid, and can be expressed using the following equation: V t = H ( D 2 + 2 DH ⁡ tan ⁡ α + 4 3 H 2 ⁡ tan 2 ⁡ α ) (2) where D is the width of the foundation, H is the embedment depth of the foundation, and α is the slope angle of the pyramidal failure surface to the vertical or to the direction of uplift load.

Based on the failure uplift resistances, T _UR, and Eqs 1, 2, the values for the slope angle α of all tests can be determined and are presented in Table 1. The magnitude of the slope angle α varies with soil properties and plate dimensions, and the values of α corresponding to the failure uplift resistances are plotted against the embedment ratio, H/D, on Figure 11 for all model tests on plate anchors embedded in UAS and CAS.

It is evident form Figure 11 that the magnitude of the slope angle α corresponding to the peak failure resistance for the foundations embedded in CAS is much larger than that for the foundations embedded in UAS. For a given mat width, D, the magnitude of the slope angle α for model plate anchors in UAS and CAS decreases with the embedment ratio, H/D. Additionally, for a given embedment ratio, H/D, the magnitude of the slope angle α for model plate anchors in UAS and CAS decreases with the plate width, D. These findings are quite important for the design of plate anchors embedded in CAS and UAS backfills subjected to uplift loads.