Performance testing and numerical modelling of ...

Performance testing and numerical modelling of reinforced soil retaining walls R.J. BATHURST, D.L. WALTERS, K. HATAMI, D.D. SAUNDERS, N. VLACHOPOULOS & G.P. BURGESS, GeoEngineering Centre at Queen's-RMC, Geotechnical Group, RMC, Kingston, Ontario, Canada T.M. ALLEN, Washington State DOT, FOSSC Materials Laboratory, Olympia, WA, USA

ABSTRACT: The paper presents selected results from a research program that is focused on the performance of full-scale geosynthetic and metallic reinforced soil retaining walls. The principal objective of the experimental work is to identify and quantify performance features of the full-scale tests. The high-quality data from physical tests is used to verify numerical model results using the program FLAC. Six walls have been tested in the experimental program, five of which were constructed with a column of drystacked modular concrete (segmental) units and one nominal identical wall constructed with a very flexible wrapped-face. All of the structures were surcharge loaded to stress levels well in excess of working load conditions. The data gathered from this program has been useful to identify sources of conservatism in current methods of analysis for reinforced soil structures in North America.

1 INTRODUCTION The Geotechnical Group of the Civil Engineering Department at the Royal Military College of Canada (RMC) is engaged in a research project related to the design, analysis, performance evaluation and numerical modelling of reinforced soil structures. To date a total of 14 full-scale retaining walls have been constructed (Bathurst 2000). Included in this number is a recent set of six carefully instrumented full-scale geosynthetic (extensible) and metallic (inextensible) reinforced soil retaining walls which are the subject of this paper. A long-term objective of the experimental and numerical work is to improve current design methodologies for reinforced soil retaining walls. The most recent test walls were 3.6-m high by 3.4-m wide by 6-m deep. Five of the walls were constructed with a column of dry-stacked modular concrete units (NCMA 1997, Bathurst and Simac 1994) and one nominal identical wall was constructed with a very flexible wrapped-face. Following construction, all of the structures were stage uniform surcharge loaded to stress levels well in excess of working load conditions. The paper presents some measured wall performance data, numerical simulation results and implications of wall performance to current design methods in North America. Additional details on this series of walls can be found in the papers by Bathurst et al. (2000, 2001). 2 EXPERIMENTAL PROGRAM 2.1 Wall configurations The full-scale walls were constructed in the RMC Retaining Wall Test Facility located within the Civil Engineering Department structures laboratory. Table 1 summarises the essential details and objectives of the six test walls completed to date in the current research program. Each of the structures following Wall 1 was constructed with one parameter changed from the control structure in order to isolate the influence of reinforcement stiffness, type, vertical spacing and facing type (i.e. hardfaced or perfectly flexible) on wall performance. Figure 1 illustrates the two different cross-sections for the walls constructed with the modular concrete facing units (segmental walls). The target setback of each layer of blocks corresponds to a facing batter of 8 degrees from the vertical. The control structure (Wall 1) was constructed with 6 layers of rein-

forcement at a 0.6 m vertical spacing and extending 2.52 m into the backfill soil. This wall was designed to satisfy current National Concrete Masonry Association guidelines (NCMA 1997). An additional design constraint was that the reinforcement layer spacing not exceed a distance equal to twice the modular block toe to heel dimension (AASHTO 1996). The same reinforcement geometry was used for Walls 2, 5 and 6 (solid horizontal lines in Figure 1). Wall 3 was constructed with 4 layers of PP geogrid at a 0.9 m vertical spacing (dashed horizontal lines in Figure 1). Wall 4 was a wrapped-face wall with 6 primary reinforcement layers of PP geogrid located at the same elevations as those shown for the control structure in Figure 1. Approximately 300 different electronic devices (instruments) were installed in each test wall to measure reinforcement strains, wall displacements, footing loads, connection loads and earth pressures. Readings were monitored continuously during construction, surcharging and excavation using an automated data acquisition system. 2.2 Materials An extruded biaxial polypropylene (PP) geogrid reinforcement was used in Walls 1 to 4. The short-term ultimate tensile strength of the PP geogrid was 14 kN/m. The aperture size of the PP geogrid was 25 mm between longitudinal members and 33 mm between transverse members. The PP geogrid used in Wall 2 was modified by removing alternating longitudinal members from samples of the original reinforcement rolls. This modified PP geogrid had a strength and stiffness that was 50% of the original geogrid and 50 mm spacing between longitudinal members. Wall 5 incorporated a knitted uniaxial PVC-coated polyester (PET) geogrid that had an ultimate strength of about 16 kN/m with aperture sizes of 27 mm between longitudinal members and 22 mm between transverse members. Wall 6 was reinforced using a welded wire mesh (WWM) with a wire diameter of 2 mm (14 gauge). The yield strength of the WWM was approximately 6 kN/m and the ultimate strength was about 7 kN/m. The aperture size for the mesh was 200 mm between longitudinal members and 100 mm between transverse members. The backfill soil used in the test walls is a clean uniform size rounded beach sand (SP) with a constant volume friction angle φ cv = 35o and a peak plane strain friction angle φ ps = 44o. The sand has a flat compaction curve and was compacted using lightweight compaction equipment. However, the first 0.5-m dis-

Table 1. Full-scale test wall program. Wall

Description

Objectives/Variables

1

Geosynthetic wall with modular block facing, using a weak biaxial polypropylene (PP) geogrid with 0.6 m spacing (CONTROL)

Reference wall for evaluating the effect of layer spacing, reinforcement stiffness and strength

2

Geosynthetic wall with modular block facing, using the same weak biaxial PP geogrid as Wall 1, but with alternating longitudinal members removed to produce a geogrid with one half stiffness and strength (0.6 m spacing)

Influence of reinforcement stiffness and strength on reinforcement loads/strains/wall deformations and overall level of safety

3

Geosynthetic wall with modular block facing, using a weak biaxial PP geogrid (0.9 m spacing)

Influence of large reinforcement spacing on reinforcement loads/strains/wall deformations and overall level of safety

4

Geosynthetic wall with a very flexible wrapped-face using a weak biaxial PP geogrid (0.6 m spacing)

Influence of facing stiffness and wall toe restraint on reinforcement loads/strains/wall deformations and overall level of safety

5

Geosynthetic wall with modular block facing, using a polyester (PET) geogrid with similar index strength to the control weak biaxial PP geogrid (0.6 m spacing)

Influence of creep deformation/behaviour on reinforcement loads/strains/wall deformations and overall level of safety

6

Generic steel welded wire mesh (WWM) (14 gauge welded wire - 2 mm diameter) reinforced wall using modular block facing (0.6 m spacing)

Influence of reinforcement stiffness on reinforcement loads/strains/wall deformations, as well as factor of safety against reinforcement rupture

Soil surface

4

6 5

3.5 Reinforcement layers

3 4 0.9 m 2

Wall 2 3 6 1 5

4.0

3.6 m

3

3.0 Elevation (m)

Facing blocks

2.5 2.0 Design batter (stepped) o 8 from vertical

1.5

2 1

0.6 m

1.0

1

0.5

0.3 m

0.0 2.52 m

Wall 1, 2, 5, 6 Wall 3

0

100

200 300 400 Distance from toe (mm)

500

600

Figure 1. Test configuration for the segmental walls.

Figure 2. Facing column profiles at the end of construction for the segmental walls.

tance directly behind the wall facing was hand tamped to the same density using a rigid steel plate. This precautionary measure was taken to minimise constructed-induced outward deformations of the blocks and reduce compaction-induced lateral stresses against the back of the facing column.

The stepped line in the figure is the target facing batter based on the geometry of the block units. The figure shows that the actual facing alignment for all of the structures is steeper than the target batter as a result of the incremental construction of the facing column together with fill placement and compaction. The relative facing profiles for Walls 1 through 3 are consistent with the expectation that the magnitude of construction-induced outward deformations of the wall face will increase for the same reinforcement type but with fewer layers or lower reinforcement stiffness. The interpretation of the relative performance of Walls 5 (PET) and 6 (WWM) with respect to the PP geogrid walls is more difficult. The larger outward deformations of the welded wire mesh wall compared to the control wall may be counter intuitive since the stiffness of the WWM material is approximately 10 times that of the control PP wall material. However, the welded wire mesh had an aperture size of 200 x 100 mm compared to 25 x 33 mm for the PP geogrid (Wall 1) and it is likely that the load transfer between the wire mesh and the sand backfill at low confining pressures was less. However, the assignment of different mechanical properties of the reinforcement materials and number of reinforcement layers to account for differences in magnitude of construction-induced movement of otherwise identical walls must be undertaken with caution. The magnitude of

3 NUMERICAL SIMULATION The dynamic finite-difference computer code FLAC (Itasca 1998) was used to simulate the results of the hard-faced walls in this test series. The numerical models used material properties for the backfill soil, reinforcement and modular block interfaces that were obtained from independent laboratory tests. For brevity only results of Wall 1 (control) are presented here. Details of numerical simulations are reported by Bathurst et al. (2001) and Hatami and Bathurst (2001). 4 MEASURED AND NUMERICAL RESULTS 4.1 Facing column profiles and displacement Figure 2 shows the results of manual survey of the facing column profiles for the segmental walls at the end of construction.

3.5 Wall 1 Measured Calculated

3.0 2.5 Elevation (m)

outward deformations is sensitive to construction technique including small variations in compaction. From a practical point of view it may be argued that despite the range of reinforcement aperture size, surface roughness, reinforcement spacing, reinforcement stiffness and global reinforcement stiffness (see Section 4.4), the maximum construction-induced wall movement varied over a narrow range of 1 to 4% of the height of the wall. This observation supports the argument that construction technique for a given facing type is an important factor controlling construction-induced wall deformations in the field. Figure 3 shows the profiles of measured and numerically calculated lateral displacements of individual facing blocks at reinforcement elevations for Wall 1 at the end of construction. The measured displacement results are readings from the potentiometers that were positioned against the facing blocks at reinforcement layer levels during construction. Accordingly, the recorded displacement values at each elevation in Figure 3 represent the magnitude of the lateral displacement of the corresponding facing block from the time of installation to the end of construction of Wall 1. The results of facing lateral displacements in Figure 3 show good agreements between recorded and calculated values for the control structure.

2.0 1.5 1.0 0.5 0.0 0

5

10

15

Xd (mm)

Figure 3. Measured and calculated (numerical) facing block lateral movement at the end of construction for Wall 1. 50 Wall 1

4.2 Horizontal and vertical toe loads

Calculated Measured

40 Toe load (kN/m)

Figure 4 shows the history of the measured and calculated (numerical) horizontal and vertical toe loads for Wall 1 during construction. The calculated history of vertical toe reaction during construction shows reasonably good agreement with the measured load data. The predicted history of horizontal toe load is slightly less accurate but captures the trend in the measured data. Superimposed on the figure is the vertical self-weight of the facing column calculated by summing the weights of each individual facing unit. The sum of the measured vertical toe loads is greater than the self-weight of the facing column. This observation is attributed to the vertical down-drag forces developed at the connections due to relative downward movement of the sand backfill directly behind the facing column. This downward movement is a result of compaction of the soil and settlement of the soil during outward rotation of the facing column. While, not shown here, the largest strains in the reinforcement layers at the end of construction were typically located at the connections which is consistent with the down-drag mechanism just described. These additional strains (loads) make it difficult to predict the magnitude and distribution of reinforcement loads in conventional design practice.

Vertical loads

30

Self-weight of facing column

20 Horizontal loads

10

0 0

5

10

15

25

30

3.5 Wall 1 Measured Calculated Coulomb

3.0

4.3 Connection loads 2.5 Elevation (m)

Figure 5 shows measured and calculated (numerical) reinforcement connection loads at the end of construction for Wall 1. The horizontal toe loads are also plotted on the figure to highlight the contribution of toe restraint to stiff facing column reactions. The data indicates that the numerical simulations were capable of capturing both the measured trend and magnitude of connection loads in the reinforcement. Also plotted on Figure 5 are the predicted connection loads at the end of construction using Coulomb lateral earth pressure theory calculated using the contributory area approach, peak plane strain friction angle of the soil and, the assumption of fullymobilised soil-wall friction angle (i.e. φ ps = δ = 44o) (NCMA 1997). In contrast to the expected triangular distribution, the measured connection loads are significantly smaller than the Coulomb predictions and more uniform with depth. The magnitude and distribution of measured connection loads in the hardfaced wall is the combined result of the rigid toe attracting a significant portion of the lateral earth forces and possible redistribution of reinforcement load during construction-induced outward movement of the facing column. Similar observations were made for the other hard-faced walls in the test program. Thus, one shortcoming of conventional earth pressure theories

20

Number of blocks Figure 4. Measured and calculated (numerical) horizontal and vertical toe loads for Wall 1 during construction.

2.0 1.5 1.0 0.5 0.0 0

2

4

6 8 Load (kN/m)

10

12

14

Figure 5. Measured, calculated (numerical) and predicted (analytical) reinforcement connection and toe loads at end of construction for Wall 1.

applied to reinforced walls with a structural facing is their inability to account for the load that is carried by the restrained toe at the base of a stiff facing column and hence is one source of conservatism in current design practice (i.e. over-estimation of

Facing column displacement (mm)

100 Wall 3 (4 layers PP/strong) 80

Wall 1 (6 layers PP/strong)

Wall 2 (6 layers PP/weak)

60 ∆

Wall 5 (6 layers PET)

40

Wall 6 (6 layers steel)

20

accounted for in current methods of design that use conventional earth pressure theories to predict reinforcement loads and hence is one source of conservatism in current design practice. The carefully instrumented and monitored walls in this ongoing research program have allowed the writers to carry out numerical simulations of the hard-faced reinforced soil walls up to the end of construction. There is generally satisfactory agreement with the measured results for wall deflections, connection loads and toe forces. At the same time, the simulation efforts highlight the difficulty to quantify factors such as minor deviations in wall construction technique, soil compaction and the very complex interactions that occur between component materials. ACKNOWLEDGEMENTS

0 0

20

40

60 80 100 Surcharge (kPa)

120

140

160

Figure 6. Overall performance of segmental walls based on postconstruction facing displacement during surcharge loading.

reinforcement loads). 4.4 Overall wall performance Figure 6 shows the overall performance of the modular block reinforced soil structures based on maximum post-construction facing displacements recorded during surcharge loading. In general, for surcharge loads less than 40 kPa, peak wall deflections increased linearly with applied load. Deflections recorded for Wall 5 (6 layers of PET) were greater than those for Wall 1 (6 layers of PP/strong) and about equal to that of Wall 2 (6 layers of PP/weak), which is consistent with the relative stiffness values of the three reinforcement products at low tensile load levels (Bathurst et al. 2001). The displacements of Wall 1 and Wall 3 (4 layers of PP) were almost identical up to the 40-kPa load level. This would indicate that for these structures, changes in reinforcement spacing had only a minimal influence on overall wall performance. Wall 6 (6 layers of WWM) generated the least facing movement which is consistent with the stiffer load-strain properties of the metallic reinforcement product. The figure also shows a highly non-linear response of the walls reinforced with the PP geogrid materials beyond the 40kPa load level. Facing deflections for Wall 2 were greater than those recorded for Wall 3, which in turn were greater than those for Wall 1. This trend is consistent with the relative global reinforcement stiffness (S g ) of the walls (i.e. S g = ΣJ/H where J is reinforcement stiffness and H is the height of the wall (Allen and Bathurst 2001)). At the 100 kPa surcharge load level, the observed peak facing displacement for Wall 5 was approximately 60% of that observed for Wall 1 (control structure) and the facing displacement for Wall 6 was only 18% that of the control structure. This is attributed to the relatively greater stiffness of the PET geogrid at higher tensile load levels and longer elapsed time and the greater stiffness of the wire mesh reinforcement (Bathurst et al. 2001). 5 CONCLUSIONS For the same reinforcement type (i.e. PP geogrid) and hard facing in this study there was a trend of increasing constructioninduced wall deformation with decreasing reinforcement stiffness and/or decreasing number of layers (i.e. decreasing global reinforcement stiffness). However, construction technique for a given facing type is also an important factor controlling construction-induced wall deformations in the field. The horizontally restrained toe of the walls in these experiments carried a significant portion of the horizontal earth forces acting on the (stiff) hard facing column. This load capacity is not

The writers would like to acknowledge the financial support of the following US State Departments of Transportation: Washington, Alaska, Arizona, California, Colorado, Idaho, Minnesota, New York, North Dakota, Oregon, Utah, and Wyoming. The writers are also grateful for the financial support of the National Concrete Masonry Association, the Reinforced Earth Company, Natural Sciences and Engineering Research Council of Canada, Academic Research Program of the Department of National Defence (Canada) and grants from the Department of Infrastructure and Environment (DND Canada). The authors would like to acknowledge the contribution of Risi Stone Systems for provision of the facing units and Terrafix Inc., Strata Systems Inc., and Modular Gabion Systems for provision of the reinforcement materials.

REFERENCES Allen, T.M. & Bathurst, R.J. 2001. Prediction of Soil Reinforcement Loads in Mechanically Stabilized Earth (MSE) Walls, Washington State Department of Transportation, Report WA-RD 522.1, 353 pp. AASHTO 1996. Standard Specifications for Highway Bridges, with 1999 Interims, American Association of State Highway and Transportation Officials, Sixteenth Edition, Washington, D.C., USA, 686 p. Bathurst, R.J. 2000. Lessons learned from full scale model tests of reinforced walls and slopes (Invited keynote paper), Proc. 2nd Asian Geosynthetics Conf., Kuala Lumpur, 29-31 May 2000: (1) 1-22. Kuala Lumpur: SEAC-IGS. Bathurst, R.J., Walters, D., Hatami, K. & Allen, T.M. 2001. Full scale performance testing and numerical modelling of reinforced soil retaining walls (Invited Keynote paper). In H. Ochiai, J. Otani, N. Yasufuku & K. Omine (eds), Landmarks in Earth Reinforcement, IS Kyushu 2001; Proc. Inter. Symp., 14-16 November 2001. (in press). Bathurst, R.J. & Simac, M.R. 1994. Geosynthetic reinforced segmental retaining wall structures in North America (Invited keynote paper), 5th Inter. Conf. Geotextiles, Geomembranes and Related Products, Singapore, 6-9 September 1994: (4) 1275-1298. Singapore: SEACIGS. Bathurst, R.J., Walters, D., Vlachopoulos, N., Burgess, P. & Allen, T.M. 2000. Full scale testing of geosynthetic reinforced walls (Invited keynote paper). In J.G. Zornberg & B.R. Christopher (eds), Geotechnical special publication No. 103, Advances in transportation and geoenvironmental systems using geosynthetics; Proc. of sessions of Geo-Denver 2000, Denver, 5-8 August 2000: 201-217. Denver: ASCE. Hatami, K. & Bathurst, R.J. 2001. Modelling static response of a segmental geosynthetic reinforced soil retaining wall using FLAC. In D. Billaux, X. Rachez, C. Detournay & R. Hart (eds), FLAC and Numerical Modeling in Geomechanics; Proc. 2nd Inter. FLAC Symp., Lyon, 29-31 October 2001: 223-231. Lisse:Balkema. Itasca Consulting Group 1998. FLAC - fast lagrangian analysis of continua (3.40). Minneapolis: Itasca Consulting Group Inc. National Concrete Masonry Association (NCMA) 1997. Design manual for segmental retaining walls, 2nd edition, second printing. Herndon: NCMA.