31
1. Introduction
Testing procedures for sampled soil have been recognized as important factors in determining soil properties. As a result, work on developing and comparing techniques for reproducing the in-situ state of soil have been
undertaken (e.g., Bjerrum 1973; Ladd and Foott 1974; Ladd and De Groot 2003;
Santagata and Germaine 2005). Most studies have focused on improving the reconsolidation stage. However, as observed by Cho et al. (2007), the effect of swelling during saturation of clay specimens causes changes in soil structure and affectes stress-strain responses at a strain of less than 0.01%. Saturation is thus considered as important as the reconsolidation stage in a triaxial test.
The saturation method commonly used in conventional triaxial tests is
applying a back pressure (ub) to compress the void air in soil specimens and make it dissolve into pore water. Simultaneously, cell pressure, slightly higher than the back pressure, is applied to prevent the bulge of the specimen which may be caused by the back pressure. The very low effective stress in the conventional saturation shows that the specimen is in the unloading state (or swelling), i.e., the path BC in Fig. 1 (where B and C denote the stress state after sampling and after conventional saturation, respectively). In this conventional saturation stage, the effective stress is almost equal to 0 and located at point C in Fig. 1. When the specimen was consolidated to the in-situ stress state by the recompression method as shown by the path CD in Fig. 1, the void ratio was changed bye1. Such a considerable void ratio change indicates that the quality of the specimen may be poor.
To lower void ratio change after recompression, the effective stress of the specimen should be maintained during saturation. This objective can be achieved by maintaining the suction in specimens at first, and then removing the suction gradually and simultaneously applying cell pressure at the same value as removed suction. This process results in the transformation of suction into positive cell
32
pressure without causing any change in the effective stress, i.e., the effective stress is still equal to the residual effective stress. The residual effective stress, namely thes in Fig. 2, is the effective stress remaining in the soil specimen after
sampling, storage, and handling (Ladd and DeGroot 2003 ; Cho et al. 2007). After that, the specimen is recompressed to in-situ stress, as indicated by point E in Fig.
1, inducing the void ratio changee2. Since the void ratio change (e2) in the path BE is significantly smaller than that due to the path CD (e1), this shows that the quality of the soil sample is not degraded in the process of saturation.
To maintain the good quality of the soil sample during the saturation stage, an apparatus capable of controlling the suction in soil specimens and connecting to the triaxial testing system should be developed. Studies related to setting up a suction control system have been done by many researchers (e.g., Cunningham et al. 2003, Jotisankasa et al. 2007). However, these suction control systems focused on studies related to unsaturated soil. The application of existing suction control systems to saturated soil tests has been limited because suction force directly applied to saturated soil will result in the decrease of degree of saturation, which is undesirable in testing on a saturated soil sample.
This paper describes the improvement of saturation in triaxial tests
considering soil suction due to sampling. An apparatus capable of suction control in triaxial tests has been developed to improve the saturation condition. Triaxial tests with and without suction control were conducted on reconstituted Taipei silty clays. The effects of saturation with suction control are verified in terms of (1) the void ratio change (e e0 ) after recompression (2) the shear modulus obtained from bender element tests during K0 consolidation, and (3) the stress-strain characteristics at small strains during undrained shearing.
2. Materials and methods
2.1 Soils
33
The soil tested was Taipei silty clay. The basic properties of the soil are shown in Table 1. The clay was dried in the oven for 24 hours and then crushed and passed through a 425-micron sieve. The reconstituted samples were created by mixing the soil powder with a water content of 93%, more than twice the liquid limit, to form slurry. The slurry was then subjected to a vertical consolidation stress of 120 kPa using 4 steps in a consolidation chamber. Use of reconstituted samples ensures that the testing results will not be affected by the variability of in-situ soils.
2.2 Development of the suction control system
Two vacuum ejectors, namely the Venturi vacuum generators, were used to produce two independent suctions required in the suction control system as shown in Fig. 3. The two suction lines were connected to the top and bottom caps, so as to control the suction in soil specimens more uniformly than using only one suction control line. Another reason for controlling the suction on the top and bottom caps separately is that different suctions on the top and bottom parts of a soil specimen may be needed for future studies. However, the controlled top and bottom suctions were identical in this study. The filters placed in the top and bottom caps were made of porous ceramic whose air entry value equals 1 bar.
Two precise electro pneumatic valves (E/P1 and E/P2 in Fig. 3) were placed on the lines in front of and controlling the vacuum ejectors. The output pressure of E/P varied with the change of input voltage, i.e., the pressure which was outputted from E/P and then went through the vacuum ejector could be controlled and adjusted simultaneously by a computer program. A designated value of suction could then be generated through the vacuum ejectors. Triangles in Fig. 4 denote the relationship between input voltages for E/P and outputted suctions. The calibrated relationships were used in controlling the suction. The error in suction controlling was less than 0.2 kPa, because the control system is automatic and feedback-controlled.
34
The negative water pressure was measured by the pressure sensors, i.e., suc1 and suc2 in Fig. 3, which ranged from -1 to 0 bars. The calibration of negative pressure sensors is shown in Fig. 4 as well and denoted by circles. Fig. 5 shows the example of suction control on a soil sample in a triaxial test, where the initial suction is 40 kPa and decreases to 0 kPa in 20 steps. Additional pressure
measurements made by mid-plane pore pressure probe (M4P in Fig. 5) and
external Druck pressure transducer (Druck in Fig. 5) were used in addition to suc1 and suc2. The rate of suction adjustment was 2 kPa/step in this test, and each step ended with the pressure reading near the step goal.
2.3 Determination of residual effective stress
The evaluation of residual effective stress can be done by measuring the suction in the soil sample. Measuring suction by undrained isotropic loading in a triaxial cell is relatively simple and convenient for the conventional triaxial testing system as discussed by Navaneethan et al. (2005). However, this technique is only reliable when the B value is very close to 1.0 (B is the Skempton’s pore water pressure parameter). For a soil sample before full saturation, the B value may be much lower than 1.0 because of the existence of air bubbles in the specimen or drainage lines. The value of suction in the soil specimen is thus expected to be overestimated significantly. The overestimated suction gives incorrect information about the residual effective stress and will result in extra isotropic consolidation on the specimen if the soil specimen is saturated at this residual effective stress.
Therefore, a direct method for measuring suction by using pressure sensors of range from -1 to 0 bar and porous ceramic filters, was adopted in this study. The limit on the suction in this study is 100 kPa due to the methods and instruments used. For measuring soil suction higher than 100 kPa, other methods are available, such as the filter paper method and high capacity tensiometers.
Performance of this measuring method will be seriously affected if the filter is not fully saturated. The initial saturation procedure for ceramic filter is similar
35
to that proposed by Ridley and Burland (1999) and Take and Bolton (2003). A chamber half full of de-aired water was prepared prior to saturating the filters. The air-dry ceramic filter was placed above the water in the chamber. Then 1 bar of vacuum was applied for one hour. Afterward, submerge the ceramic filter slowly into water while maintaining vacuum for at least 30 min. After that, vacuum was released and further time allowed for saturation of the filter under atmospheric pressure.
2.4 Experimental apparatus
A triaxial testing system, capable of performing tests at small strain and equipped for local measurements, is developed in this study. The testing system, consisting of an axial loading system, a pressure controller, and a data acquisition system, is an automated, programmable, and feedback-controlled system. A high-accuracy, direct-drive servo motor (D.D. motor) is used to provide the axial displacement in triaxial tests. As shown in Fig. 6(a), the D.D. motor (SN: NSK M-YSB 2020KN001) is fixed on the loading frame, and the transmission is an extremely precise ball screw (SN: NSK DFT 2805-5) tightened on the motor directly, i.e. no indirect transmissions such as gears or belts are used in the axial loading system.
The major advantages of a driving system with a D.D. motor are high rigidity and the absence of transmission gears. Hence, the backlash during testing was minimised to a value below the detectible limit of the displacement sensors. The resolution of the motor is 819200 steps per revolution, and every round of rotation creates 5 mm of linear motion. Thus, the minimum single axial displacement is
6.1 10 6mm (=5 819200mm), which makes it ideal for displacement-controlled tests. The maximum error due to the D.D. motor is 150 seconds in every round of rotation, i.e. the 150 seconds error will be generated in each motor rotation, i.e.
360 degrees (360 degree 60 min degree 60sec min 129200 seconds). Thus, the possible largest error in the axial displacement is 0.0116% (150 1292000),
36
or 0.58 m ( 0.0116% 5mm ) in every round of rotation of the D.D. motor.
Fig. 6(b) shows the pressure controller used in controlling the cell pressure and back pressure. The pressure controller is digital and connected to the testing system, and the resolutions for controlling and reading are 1 kPa and 0.015 kPa, respectively. This testing system uses internal measurements, including two axial and one radial local Hall effect transducers (Clayton et al. 1989), a submersible load cell (±2 kN, Sensotec), two mid-plane pore pressure probes (7 bar, GDS Instrument), and bender elements (GDS Instrument). This small strain
measurement system is capable of resolving 0.003 kPa of pore pressure, 0.001%
of axial strain, and 0.005 kPa of axial stress. The Hall effect local strain sensors, which have a linear range of ±3 mm around the electrical zero, were calibrated by a calibration rig equipped with a digital micrometer head (resolution=1 m ).
A set of bender elements was embedded in the top cap and bottom pedestal.
The shape of the received signal and uncertainties in determining the travel time of the shear wave in bender element tests were affected by near-field effects (Viggiani and Atkinson 1995; Jovicic et al. 1996; Brignoli et al. 1996). In order to understand the influence of different driven frequencies on the determination of the travel time of shear waves, bender element tests with frequencies varying from 3 kHz to 15 kHz were performed on the reconstituted Taipei silty clay at constant effective vertical stress after K0 consolidation. The methods for determining the travel time were (1) The resonance method (2) The visual inspection method, and (3) The cross-correlation method (Viggiani and Atkinson 1995; Jovicic et al. 1996;
Arulnathan et al. 1998; Lee and Santamarina 2005). The test results show that a consistent travel time or velocity can be obtained once the input frequency is higher than the resonance frequency. The disturbance of near field effect when determining the travel time is eliminated by choosing a frequency higher than the resonance frequency. Therefore a driven frequency of 10 kHz is chosen for bender element tests in this study.
37
2.5 Experimental procedure
Table 2 summarizes the triaxial tests conducted for this work. Triaxial specimens were hand-trimmed to a size of50mm×height100mm from the reconstituted samples. After mounting a specimen in the triaxial cell and installing the on-specimen sensors, the residual effective stress was measured for the
specimen tested with suction control. After suction control, the specimen was saturated at its residual effective stress. The procedure of suction control is introduced in detail as follows.
Step 1. Measure the suction in the soil specimen by a pore water pressure transducer (range from -1 to 0 bars) with a porous ceramic filter. The measured suction (s) was the initial point for controlling the suction.
Step 2. Remove the suction from the specimen gradually by a value of
s N
in each single step (where N is the number of steps).
Simultaneously, the confining pressure in the triaxial cell was also increased by the Digital Pressure Controller, in which the pressure increment was equal to s N as well. Hence, the effective stress of the specimen was kept unchanged.
Step 3. Repeat Step 2 (N-1) times. The suction in the specimen would be dissipated completely at the Nth step, and the effective stress of the specimen, equal to the residual effective stress, was now applied by the cell pressure.
Step 4. Apply a back pressure to the specimen in this step to raise the degree of saturation and simultaneously, the same amount of confining pressure should be applied until the effective stress in soil is equal to the residual effective stress.
Fig. 7 shows the variations of confining pressures, effective stresses, and pore water pressures in the specimen during the suction measurement and suction
38
control. The confining pressure was kept unchanged during the suction
measurement stage, and the effective stress in the specimen should be equal to the measured suction. After the measured suction reached equilibrium, i.e., a suction of 18.5 kPa in Fig. 7, then the suction control stage started. The suction control process was separated into 5 steps here, as shown in Fig. 7, and the adjusting suction was about 4 kPa in every single step. The cell pressure was increased simultaneously for the purpose of keeping the effective stress unchanged in each step. The effective stress in the specimen shown in Fig. 7 was always equal to the initial suction (18.5 kPa) through the whole suction control process.
For those specimens without suction control, the mean effective stresses during saturation (sat ) were set equal to 10, 5 and 2 kPa, respectively, all commonly used in the conventional saturation procedure. Each specimen was recompressed under the K0 condition to an effective vertical stress of 200 kPa or 300kPa, with full saturation (B value > 98%). Thereafter, the specimens were subjected to the axial compression under the undrained condition with a rate of 0.2
%/hr.
Bender element tests were conducted in the consolidation stage using a single-pulse sinusoidal input wave with the driven frequency determined by previous tests. The trigger points of bender element tests were at the end of every two consolidation steps. The elastic shear wave velocity, Vs, was calculated using the wave travel time determined by the visual inspection method, very close to that determined by the cross-correlation method with the chosen frequency of 10 kHz, and the tip-to-tip distance between transmitting and receiving bender elements. The shear modulus, GBE can be calculated as
2
BE s
G V (1)
where = bulk mass density of the specimen when Vs was measured.
39
3. Results of Experiments
3.1 Behaviors during K0 consolidation
Typical stress paths for triaxial tests (CK0U-AC) saturated with suction and without suction control (sat 10, 5, and 2 kPa) are shown in Fig. 8. A K0 value of approximately 0.5 is observed, reasonable for the chosen clay. The variation of
K0 with effective vertical stress for the tested clay is shown in Fig. 9. In the consolidation stage, the K0 value decreases from the initial value of 1.0 at the isotropic stress state to the end of consolidation. The final K0 values for all tests have almost converged to the value as calculated by the following equation (Mayne and Kulhawy, 1982),
sin ' 0 (1 sin ')
K OCR (2)
where ' is the effective friction angle, and OCR is the over-consolidation ratio.
The K0 value estimated by Eq. (2) is 0.53 here. Nevertheless, the K0 value for the specimen saturated with suction control is much closer to 0.53 than that of the specimen saturated without suction control when the effective vertical stress reached the pre-consolidation stress. In addition, the K0 values for the tests without suction control are all smaller than those with suction control before the effective vertical stress attained the pre-consolidation stress.
Fig. 10 shows the curve of void ratio against the logarithmic effective consolidation stress during consolidation for the tests saturated with suction control (sat 18.5kPa) and without suction control (sat 5kPa). The point A in Fig. 10 represents the stress state of soil before the sampling, i.e., at the in-situ state, where v 120 kPa. Once sampling has taken place, the stress state goes to the point B, i.e., the residual effective stress. For the conventional saturation method, i.e., when sat 5kPa, the stress state falls on the point C, with some swelling. When soil specimens were recompressed to the in-situ stress,
40
e1
(=0.063) for the test with suction control, i.e., sat 18.5kPa, was smaller than e2(=0.081) for the test without suction control, i.e., sat 5kPa. The recompression processes for both specimens saturated with and without suction control were identical. However, they resulted in different void ratio changes during recompression. It may be concluded that the quality of a specimen can be improved by the proposed suction control system.
The variation of Gvh values measured by bender elements during
recompression is shown in Fig. 11. The specimen saturated with suction control exhibited higher values of Gvh than other specimens. At the effective vertical stress equal to 147 kPa, the Gvh of the specimen saturated with suction control (sat 18.5kPa) is 14% higher than that saturated without suction control (sat 5kPa).
3.2 Stress-strain behaviors during undrained shearing
After K0 consolidation, all the specimens were subjected to undrained axial compression. Fig. 12 illustrates the differences in the initial part of the
stress-strain responses by plotting the curves of Young’s modulus normalized to the undrained shear strength. Although all degradation curves of stiffness are similar, the stiffness of the specimen saturated at sat 2kPa at small strain, i.e., strain equals 10 %3 , is about 20% lower than that saturated with suction control.
Results from the comparison of the stiffness at small strain indicate that the conventional saturation method reduces the initial stiffness of soils, and the saturation method with suction control is helpful for keeping sample quality.
The undrained failure characteristics for reconstituted clay for different sat are summarized as listed in Table 2. Unlike the stress-strain behavior at small strain level, all the tests exhibite the same value of normalized undrained shear strength, i.e., Su vc 0.3. This may imply that the undrained failure
characteristics were little affected by saturation procedure.
41
4. Discussion
Both the suction measurement and vacuum source used here were in a range of -1 to 0 bars. The applicability of the proposed method of specimen saturation in triaxial tests is limited to specimens with suction less than 100 kPa from the measurement point of view and the control point of view. For specimens whose suctions are less than 100 kPa, the proposed saturation method is effective in improving the quality of specimens as discussed in the following.
The sample quality can be evaluated by either the void ratio change during recompression or the shear modulus measured by bender elements. The void ratio changes during recompression for both tests saturated with and without suction
The sample quality can be evaluated by either the void ratio change during recompression or the shear modulus measured by bender elements. The void ratio changes during recompression for both tests saturated with and without suction