Vous êtes sur la page 1sur 10

Study of Geotechnical Failures through Physical Modeling

C. F. Leung, D. E. L. Ong & Y. K. Chow


Centre for Soft Ground Engineering, Department of Civil Engineering, National University of Singapore, Singapore 117576

ABSTRACT: The use and benefits of physical modeling in the study of geotechnical failures is demonstrated in this paper. Centrifuge model studies on the performance of piles behind an unstable retaining wall in clay as well
as the responses of piles due to tunnel collapse in sand are used as illustrative cases. An enhanced image processing system is employed to visualize the soil movement patterns and trends during the failure of the wall or the tunnel. The observations together with the measured bending moment and deflection profiles along the pile and ground settlements provide valuable information on the failure mechanism of the problems understudy without the occurrence of actual failures in the field. Prediction of pile responses using a numerical method developed at the National University of Singapore is also presented in this paper to evaluate the reliability of the centrifuge test observations and results.

Keywords: Soil flow, failure, tension cracks, soil movement, limiting soil pressure, image processing 1 INTRODUCTION Geotechnical failures often occur when sub-standard construction procedures have been adopted. High variability and insufficient or inaccurate soil investigation results regarding subsurface soil layers and properties can also lead to erroneous designs. Many geotechnical failures have been reported. For instance, Ting et al. (1994) reported the failure of piles supporting an embankment due to landslip, Poulos (1994) reported the piled foundation failure of an office tower due to a nearby excavation, and the failure of a pile-supported wharf structure due to riverbank movement has been described by Ting and Tan (1997). When failures occur, lives, money and confidence are lost. Painful lessons can be learnt from each failure so that such failures can be avoided in the future. Nevertheless, it is also important to understand the behaviour of the structures during and after failure so as to widen the knowledge beyond the serviceability limits of structures. Reallife structures are not normally built and then tested to failure as tremendous amount of money and time is involved. Therefore, full-scale testing of a structure to failure is generally deemed to be uneconomical and undesirable. As such, physical modeling is considered an attractive alternative to study geotechnical failures. Leung et al. (2000) reported the results of a centrifuge model study on the behaviour of piles behind a collapsed retaining wall in sand. In this paper, the responses of piles behind an unstable retaining wall in clay and adjacent to a tunnel that subsequently collapses upon excavation are used as illustrative examples to demonstrate the benefits of physical modeling in the study of geotechnical failures. Centrifuge model tests can be carried out under a controlled environment where the soil strength profiles, soil deformation and elapsed time can be measured with reasonable accuracy resulting in reliable test results. Besides that, the consistent repeatability of centrifuge experiments also renders centrifuge model study attractive and relatively economical. 2 EXPERIMENT SET-UP 2.1 Model pile and retaining wall The first study involves the investigation of pile responses behind a retaining wall that collapses upon reaching certain excavation depth. Figure 1 shows the centrifuge model setup. All the experiments were conducted at 50g on the National University of Singapore geotechnical centrifuge. The model pile was fabricated from a hollow square aluminium tube with an outer diameter of 9.53 mm and a wall thickness of 3.18 mm. Ten pairs of strain gauges were attached to the opposing faces of the model pile at vertical intervals of 25 mm. The

final width of the pile is 12.6 mm. At 50g, the equivalent pile width is 630 mm in prototype scale. The total length of the pile is 350 mm (17.5 m in prototype scale). The pile has an embedment depth of 250 mm (12.5 m) into the clay. A thin layer of epoxy was applied to the entire pile length to ensure that the strain gauges and all connections were waterproof and well protected. The stiffness of the epoxy coating is assumed to be negligible when compared to that of the aluminium model pile. The prototype bending rigidity, EI, of the model pile is approximately 2.2 x 105 kNm2. This is somewhat equivalent to a 12.7 mm thick, 610-mm diameter steel pipe pile in the field. The model retaining wall is fabricated using a 3 mm (150 mm in prototype dimension) thick aluminum plate. Its prototype bending rigidity, EI, is 24 x 103 kNm2/m, which is equivalent to that of a FSP II A sheet pile.

Table 2. Physical properties of sand (after Ong et al., 2003a) Mean grain size Uniformity coefficient Specific gravity, Gs Friction angle (50-100 kPa) 0.16 mm 1.3 2.65 43o

2.3 Test preparation and procedure Kaolin powder was mixed with water at a water content of 120% in a de-airing mixer. The slurry was then placed in the model container under water. Subsequently, a 17-kg plate was placed on top of the slurry to stiffen it. Then, the sample was placed on a loading frame for 1-D consolidation under a load of 20 kPa for 3 days. After that, self-weight consolidation of clay was carried out under 50g for about 6 to 7 hours. The ground surface settlements were monitored by displacement transducers (LDVT). After about 90% consolidation had been reached, the centrifuge was spun down. Pore pressure transducers (PPT) were then embedded at positions shown in Figure 1. Subsequently, the model wall and pile were jacked vertically into place using a guide at 1g. Excavation was then carried out. The excavated clay was replaced by ZnCl2 solution contained in a latex bag. The density (16.5 kN/m3) and height of the ZnCl2 solution were made to be identical to those of the clay that had been excavated. After that, the front perspex face of the container was removed so that soil movement markers could be placed on the clay at 20 mm square grids. LVDTs were installed to measure the ground settlements behind the excavation. The pile head deflection, during and after excavation was monitored by two non-contact laser displacement transducers. A highresolution image-processing camera was mounted in front of the perspex window of the model container. The camera is capable of recording a high-resolution image with a pixel-to-pixel spacing of less than 0.1 mm. Finally, the model package was put together and spun up to 50g for reconsolidation. When both pore water pressures and ground settlements behind the wall showed negligible changes, the ZnCl2 solution was then released to depict excavation at 50g. In prototype scale, the simulated excavation rate was about 0.56 m per day, which is equivalent to normal real-life average rate of excavation. The stress history of the clay sample was intended to simulate that of a normally consolidated clay deposit with a 2.8-m overconsolidated crust. The detail test procedures are described in Ong et al. (2003a). Two tests depicting the collapse of retaining wall were conducted in the present study. The configurations of the tests are shown in Table 3.

Figure 1. Centrifuge model set-up (all dimensions in mm) (after Ong et al., 2003a)

2.2 Soil properties The properties of the clay and sand used are shown in Tables 1 and 2, respectively. The undrained soil strength profiles measured by inflight bar penetrometer tests are given in Ong et al. (2003b).
Table 1. Physical properties of kaolin clay (after Ong et al., 2003a) Specific gravity, Gs Liquid limit, LL Plastic limit, PL Compression index, Cc Swelling index, Cs Void ratio at 115 kPa at NC line Permeability at 115 kPa at NC line 2.6 80 % 40 % 0.65 0.14 1.67 1.3x10-8 m/s

Table 3. Configuration of Tests 1 and 2 Item Distance of pile from wall Excavation depth Clay thickness Sand thickness Wall toe embedment in sand Test 1 3m 1.8 m 10.5 m 2.0 m 0 m; Floating in clay Test 2 3m 2.8 m 6.5 m 6.0 m 1.5 m

also located at 8.75 m below the ground at an excavation depth of 1.2 m. Subsequently, this value drops slightly with increasing excavation depth but somehow increases again after the excavation depth exceeds 2.0 m, as shown in Figure 3. A maximum pile bending moment of 250.7 kNm is recorded at the end of excavation of 2.8 m.

Bending moment (kNm)


0 0
Symbol Excavation depth (m)

100

200

300

3 TEST RESULTS AND DISCUSSIONS The pile behaviour in Tests 1 and 2 are discussed with particular reference to the measured soil deformation patterns and the pore water pressure responses. 3.1 Pile response The bending moment profiles along the pile at different excavation depths for Test 1 are shown in Figure 2. The maximum pile bending moment is located at 8.75 m below the ground level. A maximum value of 235.7 kNm is recorded at an excavation depth of 1.2 m. As excavation continues to its final depth of 1.8 m, the bending moment is noted to reduce to 185.8 kNm.
2.5

0.4 0.8 1.2 1.6 2.0 2.4 2.8

Depth (m)

5 7.5 10 12.5

Figure 3. Development of pile bending moment profile over time (Test 2)

Bending moment (kNm)


0 2.5 0 50
Symbol

100

150

200

250

3.2 Wall and soil deformation


3

Wall head lateral movement (m)

Excavation depth (m)

1.0 1.2 1.4 1.6 1.8

2.5 2 1.5 1 0.5 0


Test 1 Test 2
Settlement Settlement Wall head 1.5 m 4.5 m movement behind wall behind wall

Depth (m)

5 7.5 10 12.5

Ground settlement (m)

0.3 0.6 0.9 1.2 1.5 1.8 0.1 1 10 100 1000

Figure 2. Development of pile bending moment profile over time (Test 1)

Figure 3 shows the development of bending moment profiles along the pile for Test 2 in which greater excavation depth was involved. During the early excavation stages, the measured pile bending moments show a similar trend as that in Test 1. The maximum pile bending moment is about 186.7 kNm

Time from start of excavation (days)


Figure 4. Soil and wall deformation over time

-1

Soil movement (mm) 1000

Depth (m)

-3

75 65 55 45 35 25

-5

Excavation 0.6 m, -7 1.0 days


-1

15 5

300 250

Depth (m)

-3

200 150 100 50

-5

Excavation 1.0 m, -7 1.7 days

-1
500 450 400 350 300 250 200 150 100 50

-5

Excavation 1.4 m, -7 2.4 days

-1

Depth (m)

-3

750 650

Depth (m)

-3

550 450 350 250 150 50

-5

Excavation 1.8 m, 3.0 days -7


1 3 5 7

Distance (m)

Figure 5. Selected soil movement vectorial plots for Test 1

Figure 4 shows the wall head movements and ground settlements over time for both tests. As expected, the wall and ground deformations in Test 2 are larger than those in Test 1. Both the ground settlement and wall deflection follow a similar pattern over time. For both Tests 1 and 2, when excavation depth exceeds 0.6 m, the image processor shows that the clay surface starts to move past the pile head. When the excavation depth exceeds about 0.8 m, it is observed from the image processor that some

fissures develop at the ground surface in Tests 1 and 2. Water can hence flow freely into the increasing number of fissures as excavation progresses. The development of these fissures further reduces the stability of the wall, thus causing the wall to tilt and the clay to be sheared progressively over time. Therefore, the overall clay stiffness is reduced. The observed slip lines, which shows the extent of the deformed zone around the pile, is similar to the characteristics meshes postulated by Randolph et al. (1984) for a plastically deforming cohesive soil.

High-resolution pictures were taken during various excavation stages of the tests, as shown in the left-hand side photos of Figure 5. It is evident that tension cracks have developed when the excavation depth exceeds 1.0 m. These cracks cause the loss of contact of clay in front of the pile. This in turn prevents the full transmission of soil pressure onto the pile. It is also probable that the soil would flow past the pile and not to have exerted full pressure on the pile. Figure 6 shows the top plan view of the deformed soil around the pile after excavation. It is again evident that there is a considerable drop in soil-pile contact and this might help to explain the reduction in pile bending moments after the development of tension cracks. Nevertheless, the pile head deflection remains fairly constant (see Figure 7) as the pile is continuously being pushed by the separated soil mass behind the pile. However, such pile behaviour is only noted prior to the occurrence of a fully developed active shear failure wedge as in Test 1.

plotted in terms of soil displacement vectors and shown in Figure 8. It clearly shows that the sudden changes in the length of vectors reflect the development and occurrence of the active shear failure wedge.
250 250 200 150
Deflection

Bending moment (kNm)

200 150 100 50 0 0 0.4 0.8

100 50 0 1.2 1.6 2

Excavation depth (m)


Deformed zone Large tension crack
Figure 7. deflection Development of pile bending moment and

Soil flow direction

The development of the active shear failure wedge in front of the pile causes the pile bending moment to drop, similar to that observed in Test 1. However, the mass of soil behind the pile starts to move forward in response to the increasing excavation depth, thus causing the pile bending moment to increase again. This occurs when the excavation depth exceeds 2.0 m as shown in Figure 3. 3.3 Pore water pressure The variations of excess pore water pressure with time for Tests 1 and 2 are shown in Figures 9 and 10, respectively. All PPTs register excess pore water pressures immediately after excavation. Nevertheless, PPT 1 of Test 1 registers positive excess pore water pressures shortly after the completion of excavation. This is due to the water level after excavation being higher than its hydrostatic level caused by ground surface settlements. Besides that, the flooded tension crack also caused the free water and air to enter PPT 1. Owing to relatively greater soil deformations near the wall, PPT 2 registers a greater dissipation of excess negative pore water pressures as opposed to PPT 3, which is located further away from the wall. The difference in excess pore water pressures creates a hydraulic gradient that leads to pore water pressure redistribution and hence, the reconsolidation of the clay over time. This is somewhat analogous to that observed by Stewart (1992).

Deformed zone

Figure 6. Soil flow and tension cracks around the pile

Therefore, Test 2 can be used to assess the pile behaviour during and after the occurrence of a fully developed active shear failure since the excavation depth is large enough for the shear failure wedge to develop. When excavation exceeds 1.8 m, the photos show that a shear band appears gradually. As excavation progresses further, an active shear failure wedge has initiated, starting from the location of the tension cracks and gradually moves down until it intersects the wall toe. As the wall is not braced, it is observed to have rotated around the pivot, which is located at the clay-sand interface. These observations are consistent to the findings of Bolton and Powrie (1987) and Wei (1997). The soil movement immediately after excavation and the development of the active shear failure wedge are

Deflection (mm)

Bending moment

-4

-4

-6 -5 -3 -1

-6 1 3 5
Soil movement (mm) 1800

Distance (m)

Figure 8

Development of active shear failure wedge during wall collapse (Test 2)

Excess pore water pressure (kPa)

5 0

PPT 1

PPT 2

Test 1 because the wall in Test 1 is embedded entirely in clay, whose permeability is much lower and hence, the dissipation of excess negative pore water pressure is not as easy and obvious as compared to Test 2, where the wall is embedded into the sand layer so that seepage can occur easier.

PPT 3

Excess pore water pressure (kPa)

-5 -10 -15 -20 -25

0 PPT 1 -10 PPT 3 PPT 2 -20 PPT 4 -30

PPT 4

0 50 100 150 200 250 300 350 Time after start of excavation (days)
Figure 9. (Test 1) Excess pore water pressure variation over time

After excavation, the ground water level at the excavated side will drop. This creates a water pressure head difference between the retained and excavated sides. PPT 4 in Test 2 shows a rebound or dissipation of excess negative pore water pressure over time. However, a rebound is not observed in

-40 0 50 100 150 200 250 300

Time after start of excavation (days)


Figure 10. (Test 2) Excess pore water pressure variation over time

Depth (m)

-2

0 2.5

Uncorrected py = 9cu
(a) (b)

Depth (m)

5 7.5 10 12.5 0 2.5


Excavation depth (m) 0.6 0.8 1.0 1.2 1.4 Measured Predicted

Depth (m)

5 7.5 10 12.5 0 100 200

Corrected py = 3cu

(c)

(d)

300

400

500 -20

20

40

60

80 100 120

Bending moment (kNm)


Figure 11.

Deflection (mm)

Measured and predicted bending moment and deflection of pile for Test 1 (after Ong et al. 200b)

Since the wall in Test 1 is floating in clay without embedding into the stiffer sand layer, the reconsolidation of soil caused by pore water pressure redistribution becomes more dominant than the seepage caused by the head difference after excavation, as revealed by PPT 4 readings in Test 2 shown in Figure 10. The excess negative pore water pressures in the retained side for Tests 1 and 2 are relatively small as they may be partly cancelled out by the positive pore water pressures generated by the undrained shearing of the clay. Similar observations were also reported by Kimura et al. (1994). 4 NUMERICAL ANALYSIS The numerical method developed by Chow and Yong (1996) is used to back-analyze the responses of a single pile due to excavation-induced soil movement obtained from the centrifuge tests. This numerical method has been used successfully by Leung et al. (2000) to back-analyze the pile

responses due to excavation-induced soil movement in sand. The concept of analysis is based on finite element method where the pile is represented by beam elements and the soil is idealized using the modulus of subgrade reaction. The non-linearity of the soil behaviour can be incorporated to an extent by limiting the soil pressure that could act on the pile. The numerical analysis requires the knowledge of the pile flexural rigidity (EI), the distribution of lateral soil stiffness (Kh) with depth, the limiting soil pressure (py) that acts on the pile and the lateral soil movements. This approach is used in the present study to predict the pile responses in Test 1. 4.1
Input soil properties

The distribution of lateral soil stiffness with depth (Kh), Youngs modulus of the soil (Es) as well as the soil resistance or limiting pressure (py) in kaolin clay are described in detail in Ong et al. (2003b).

4.2

Prediction and discussions

When the excavation depth exceeds 1 m in Test 1, the predicted pile responses are grossly overestimated, as shown in Figures 11a and b. This is because when the clay starts to yield, the limiting soil pressure may have been reached The yielding and/or failure behaviour of the clay is evidently demonstrated by the bar penetrometer test results (Ong et al., 2003b), where the undrained shear strength (cu) of the clay reduces significantly after the soil has been excavated. By performing backanalysis of the centrifuge results, reasonable predictions of the pile responses can be obtained by adopting py = 3cu (Ong et al., 2003b). Subsequently, the corrected predicted pile responses are shown in Figures 11c and d, which reveal considerably better agreements with the measured pile responses. 5 TUNNEL-SOIL-PILE INTERACTION The second example involves the investigation of responses of piles close to a collapsed tunnel. Feng et al. (2002) presented results of centrifuge model study to investigate pile responses prior, during and after the collapse of a model tunnel. A polystyrene foam core, which is placed inside a brass foil-made model tunnel lining, is dissolved by an organic solvent during centrifuge flight to simulate the process of tunnel excavation. Strain gauges attached on the model pile shaft are used to measure the bending moment and axial load profiles along the pile during tunnel excavation.

All the centrifuge tests were conducted at 100g. In one test, the brass foil is 0.05mm thick simulating a relatively weak tunnel lining, and wrapped around the tunnel-shaped polystyrene foam core. The lining was soldered using tin solder and an electronic soldering gun. The tunnel is 210mm long (21m at prototype scale) and 60mm (6m) in diameter. Figure 12 shows the centrifuge model setup and Table 4 lists the test parameters in prototype scale.
Table 4. Test parameters (after Feng et al., 2002) Tunnel diameter Tunnel depth Ground depth Ground type Distance of pile from tunnel vertical center-line 6m 16 m 15 m Dry sand 9m

The weak model tunnel lining is not strong enough to support the soil and the tunnel completely collapses during excavation. This test condition represents the worst situation in practice. The first pile, which is used to measure the bending moment profile, was positioned 90mm (9m in prototype scale) from the tunnel vertical centerline. The pile has free head and tip conditions as well as a bending capacity of about 3000 kNm. Figure 13 shows the maximum induced bending moment profile of the pile due to the tunneling process. The maximum bending moment is 1345 kNm located approximately at the depth of horizontal tunnel center. As expected, the bending moment at pile head and pile tip is zero.

Model container
0 0
LVDT

Bending moment (kNm)


500 1000 1500

5 10 15 20 25

strain gauge

sand tunnel

Depth (m)

pile

Figure 12. al., 2002)

Centrifuge model set up for tunnel (after Feng et

Figure 13. al., 2002)

Bending moment along pile shaft (after Feng et

The second pile used to measure the axial load profile was positioned on the opposite side of the tunnel having the same distance from the tunnel vertical centerline as the first pile. However, this is a floating pile with its tip at 20 mm above the container base, and its axial load capacity is about 2500 kN. Figure 14 shows the induced axial load increased downwards from the pile head, and reached a maximum value of 1232 kN at approximately the depth of tunnel center, and then reduced towards the pile tip. The soil movements were tracked by the movement of the beads placed on the sand before the test. It is noted that the soil above the tunnel had moved downwards, while the soil below the tunnel had moved upwards during the tunneling process as shown in Figure 15.

Axial load (kN)


Figure 15. Soil movement due to tunneling

0 0 5

500

1000

1500
6 CONCLUSION In this paper, the benefits of physical modeling in the study of geotechnical failures are demonstrated by means of examples using studies on pile responses behind an unstable retaining wall and adjacent to a tunnel that collapses upon excavation. Excavation of soil causes lateral stress relief due to the removal of overburden pressure. Therefore, the strength and stiffness of the clay are reduced after an excavation. Such reduction in soil strength may have contributed to the soil flow phenomenon and the development of tension cracks, which subsequently affect the behaviour of piles embedded in soil undergoing large deformation. The deformation of clay can be correctly modeled by the numerical analysis if appropriate limiting soil pressure values are used. This can be achieved by performing inflight bar penetrometer tests to obtain the clay undrained shear strength profile before and after excavation. Tunnel-soil-pile interaction has also been studied. Owing to the collapse of tunnel lining, it is observed that about half of the ultimate bending and axial load capacities of the pile have been induced on an adjacent pile from the present centrifuge model study. This would reduce the capability of the installed piles to sustain the serviceability loads of the structures. Ground subsidence due to the collapse of tunnel lining also poses a great threat to existing adjacent buildings, especially those founded on pad footings, where differential settlement will cause structural distress. The two studies demonstrate that centrifuge modeling can be a versatile tool to study geotechnical failures as reliable results can be produced with reasonable accuracy. The

Depth (m)

10 15 20 25

Figure 14. al., 2002)

Axial load of pile due to tunneling (after Feng et

This helps to explain why the axial load transfer of the pile below the tunnel elevation is positive. Compared with the soil movements above the tunnel, the soil movements below the tunnel were generally much smaller. The ground surface settlements due to tunneling were measured with an array of LVDTs. The observed ground surface settlement trough is noted to resemble the classical Gaussian-shape settlement trough. From this study, it can be observed that the maximum bending moment and axial load of the pile occurred at approximately the depth of the tunnel horizontal centre-line. However, the axial load gradually reduced below the tunnel centre-line. The induced pile bending moment and axial load were approximately 55% and 50% of the bending and axial capacities of the piles, respectively.

observations and results facilitate a much better understanding of the behaviour of piles due to failure of adjacent geotechnical structures. ACKNOWLEDGEMENTS The authors wish to acknowledge the contributions of Mr S-H Feng on the study of tunnel-soil-pile interaction study. The assistance of Geotechnical Centrifuge Laboratory professional and technical staff is also gratefully appreciated. REFERENCES
Bolton, M. D. and Powrie, W., The Collapse of Diaphragm Walls Retaining Clay, Geotechnique, Vol. 37, No. 3, pp. 335-353, 1987. Chow, Y. K. and Yong, K. Y., Analysis of Piles subject to Lateral Soil Movements. Journal of The Institution of Engineers, Singapore, Vol. 36, No. 2, pp. 43-49, 1996. Feng, S. H., Leung, C. F., Chow, Y. K. and Dasari, R., Centrifuge Modelling of Pile Responses due to Tunneling, 15th KKCNN Symposium on Civil Engineering, Singapore, 2002. Kimura, T., Takemura, J., Hiro-oka, A., Okamura, M. and Park J., Excavation in Soft Clay Using an In-flight Excavator, Centrifuge 94, Leung, Lee and Tan (eds), pp. 649-654, 1994. Leung, C. F., Chow, Y. K. and Shen, R. F., Behaviour of Pile subject to Excavation-induced Soil Movement, Journal of Geotechnical and Geoenvironmental Engineering, Vol. 126, No. 11, pp 947-954, 2000. Ong, D. E. L., Leung, C. F. and Chow, Y. K., Time-dependent Pile Behaviour due to Excavation-induced Soil Movement in Clay, Proc. 12th Pan-American Conference on Soil Mechanics and Geotechnical Engineering, MIT, Boston, U.S.A, Vol. 2, pp. 2035-2040, 2003a. Ong, D. E. L., Leung, C. F. and Chow, Y. K., Piles subject to Excavation-induced Soil Movement in Clay, In publication, XIIIth European Conference on Soil Mechanics and Geotechnical Engineering, Prague, Czech Republic, 2003b. Poulos, H. G., Design of Piles subjected to Lateral Soil Movements. 5th Indonesian National Geotechnical Conference. Randolph, M. F. and Houlsby, G. T., The Limiting Pressure on a Circular Pile Loaded Laterally in Cohesive Soil, Geotechnique Vol. 34, No. 4, pp. 613-623, 1984. Ting, W. H., Chan, S. F. and Ooi, T.A., Design Methodology and Experiences with Pile Supported Embankments, Symposium on Development in Geotechnical Engineering, AIT, Thailand, 1994. Ting, W. H., Tan, Y. K., The Movement of a Wharf Structure subject to Fluctuation of Water Level, Proc. of XIVth Int. Conf. on Soil Mechanics and Foundation Engineering, Hamburg, 1997. Wei, J., Centrifuge Modelling of Deep Excavations., M.Eng Thesis, National University of Singapore, 1997.

Vous aimerez peut-être aussi