Structural Behavior of a Pile-Supported Embankment

Structural Behavior of a Pile-Supported Embankment G. S. Wachman1; L. Biolzi2; and J. F. Labuz, F.ASCE3 Abstract: The stress field in a pile-supported 3.9-m-high embankment was interpreted through three-dimensional finite-element modeling, and evaluated by field measurements involving strain gauges on the piles and earth pressure cells at the top and bottom of a 0.9-m-thick geogrid-reinforced platform. Analyses of the numerical results and the experimental data suggest that a vaultlike arch developed within the embankment, such that the vertical stress at the top of the platform was concentrated above the piles and virtually no vertical stress was measured between the piles. A similar situation existed within the platform, where an almost stress-free region between the piles was experimentally detected and numerically verified. From a structural point of view, a supporting skeleton was formed from a pile extension through the platform, a type of stress diffusion problem, and an arching effect appeared mainly in the embankment due to the very large stiffness of the piles in comparison to the surrounding media. DOI: 10.1061/共ASCE兲GT.1943-5606.0000180 CE Database subject headings: Structural behavior; Earth pressure; Embankment; Field tests; Geogrids; Load transfer; Piles; Platforms; Stress distribution. Author keywords: Earth pressure; Embankment stability; Field tests; Geogrids; Load transfer; Piles; Platforms; Stress distribution.

Introduction A number of techniques have been developed in order to deal with anticipated settlements in soft soils below embankments. An option gaining in popularity is to build the embankment on a grid of piles or columns that are driven or constructed to a more competent underlying layer, such as bedrock. Assuming the embankment load is transferred almost entirely to the piles, and thus the competent layer, the problems associated with the soft soils can be avoided. The mechanism that allows that load transfer is soil arching 共Hewlett and Randolph 1988兲. It has been found that the addition of one or more geogrid layers at the base of the embankment, just above the piles, facilitates the transfer of the embankment load to the piles, and allows for greater pile spacing. Lack of a standard design has led to the development of many design theories and methods 共c.f. Love and Milligan 2003; van Eekelen et al. 2003,2008; Collin 2004兲. Additionally, the interaction between the embankment, geogrid, and columns is complex, and much remains unknown about the precise mechanism by which the load transfer occurs. The objective of this research was to evaluate the performance of a pile-supported embankment constructed as part of the expansion of a highway. Fig. 1 shows a schematic cross section of the project. The design consisted of 207, 0.3 m 共1 ft兲 diameter, and 1 Civil Engineer, U.S. Army Corps of Engineers, 190 5th St. East, Suite 401, St. Paul, MN 55101 共corresponding author兲. E-mail: greg. [email protected] 2 Professor, Dipartimento di Ingegneria Strutturale, Politecnico di Milano, 20133 Milano, Italy. 3 MSES/Miles Kersten Professor, Dept. of Civil Engineering, Univ. of Minnesota, Minneapolis, MN 55455. Note. This manuscript was submitted on November 24, 2008; approved on June 15, 2009; published online on June 20, 2009. Discussion period open until June 1, 2010; separate discussions must be submitted for individual papers. This paper is part of the Journal of Geotechnical and Geoenvironmental Engineering, Vol. 136, No. 1, January 1, 2010. ©ASCE, ISSN 1090-0241/2010/1-26–34/$25.00.

6.4 mm 共0.25 in.兲 thick steel pipe piles on a triangular grid with center-to-center spacing s = 2.1 m 共7 ft兲 driven to bedrock, and circular steel pile caps with diameter d = 0.6 m 共2 ft兲; the piles at the edge of a reinforced slope were battered. The 3.9 m 共13 ft兲 high embankment used a 0.9 m 共3 ft兲 thick load transfer platform 共LTP兲, which consisted of three layers of uniaxial geogrid 共Synteen SF 80兲 separated by layers of highly frictional 共gravel兲 fill, followed by 3.0 m 共10 ft兲 of sandy soil. Of particular interest were the following quantities: 1. The portion of embankment load carried by the piles. 2. Earth pressure between the piles at the base of the LTP. 3. Strain in the geogrid layers within the LTP. 4. Earth pressure at the top of the LTP, directly above and between piles. In order to investigate these quantities, a field monitoring program was designed and implemented, consisting of 48 sensors: strain gauges, earth pressure cells 共EPCs兲, and settlement systems 共Wachman and Labuz 2008兲. The latter instruments were affected by temperature and those data will not be discussed.

Literature Review Rathmayer 共1975兲 reported on three test embankments, which were instrumented with EPCs and load cells in order to investigate 共a兲 transfer of the embankment load to the piles; 共b兲 vertical stress distribution over a pile cap; and 共c兲 the effect of pile cap shape on loading over the pile cap. The stress distribution above the pile cap was found to be highly nonuniform, concentrated largely on the outer ring of the cap. Reid and Buchanan 共1984兲 detailed the use of pile-supported embankments for bridge approaches. It was noted that in these embankments a membrane layer was placed directly above the pile caps in order to minimize settlement in the areas between pile caps and to encourage arching. Instrumentation at two separate sites included an array of EPCs above pile caps, piezometers, inclinometer tubes, settlement profile tubes, and pile load cells. Fluet et al. 共1986兲 reported on a

26 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JANUARY 2010

Downloaded 24 Dec 2009 to 131.175.24.250. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright

granular fill 3.0 m high

pile-supported embankment 3.9 m high

surcharge, lightweight fill organic soils

load transfer platform 0.9 m thick pipe piles original profile geofoam (lightweight fill)

to promote a basis for design. A limit equilibrium analysis of vaulted arches was guided by model tests, and the work is notable for capturing the essence of the structural behavior. Russell and Pierpoint 共1997兲 organized the results of Hewlett and Randolph into expressions for design.

sands and loams

Description of Site sandstone bedrock

Fig. 1. Schematic of the foundation system showing the pilesupported embankment

test embankment spanning two supports that was constructed to investigate the effects of a geogrid layer placed at the base of the embankment. The foundation soil between the supports was modeled using inflatable air bags in which the pressure could be controlled. Gartung et al. 共1996兲 reported on a railway embankment reinforced with three layers of geogrid supported by a rectangular grid of piles. Strain on geogrid was measured to be between 0.3 and 1%, with the greatest value taking place at the central location between four piles. Maddison et al. 共1996兲 described the design and performance of an embankment supporting a new toll plaza. Vibro concrete columns were constructed in a triangular grid pattern to support a LTP containing two layers of biaxial geogrid. Construction monitoring was performed using a series of hydraulic pressure cells and settlement markers. Stewart et al. 共2004兲 evaluated the performance of a test embankment at the interchange between I-95 and U.S. Route 1 in the State of Virginia. Dry deep mixing was used to produce 59 columns, which gave vertical support to a test embankment laterally supported by a geosynthetic-reinforced mechanically stabilized earth wall on one side and a 2H:1V slope on the other. Field measurements of vertical stress directly above columns and between the columns were taken and used to calibrate a numerical model. Jones et al. 共1990兲 quantitatively considered the effects of geotextile reinforcement through a two-dimensional 共2D兲 finiteelement 共FE兲 analysis to obtain a more accurate prediction of tensile forces in the reinforcement. Parameters that were studied were material properties 共friction, cohesion, unit weight, Poisson’s ratio, and modular stiffness兲 associated with the fill as well as the foundation material, embankment height, pile spacing, and pile cap diameter 共combined to provide roughly 10% areal coverage by pile caps兲. Low et al. 共1994兲 performed model tests and an analysis of a pile-supported embankment with geotextile reinforcement, but with the addition of cap beams spanning the piles in the direction perpendicular to the longitudinal axis of the embankment. Design parameters of efficacy, competency, and stress reduction ratio were calculated given the embankment height and beam spacing. Han and Gabr 共2002兲 used numerical modeling to analyze the case of a square pile layout, studying total and differential settlements, vertical stresses above and below the geosynthetic, soil arching between piles, stress concentration above piles, and tension in a geosynthetic layer. Results of the cases of reinforced and unreinforced embankments showed the effectiveness of geogrid in minimizing total and differential settlements. van Eekelen et al. 共2008兲 presented a study of British and German standards for piled embankments using experimental data and numerical modeling. Two years of field measurements are compared with predictions. A seminal study was performed by Hewlett and Randolph 共1988兲, where both experimental and analytical models were used

The site is located 65 km 共40 mi兲 northwest from St. Paul, Minnesota on Trunk Highway 共TH兲 241, about 600 m 共2,000 ft兲 southwest of the I-94/TH 241 interchange, involving an expansion of the highway from two to four lanes. The section of TH 241 for which the pile-supported embankment was chosen was located between stations 167+ 50 and 171+ 00. The original two-lane highway was built in 1938 and overhauled in 1956 and 1973, with several bituminous resurfacings since the original construction. No measures were taken to deal with the deep soft soils encountered at the project site prior to the expansion. Evidence of consolidation due to the initial embankment load could be seen prior to the 2006 expansion; for example, a depression in the road was visible across the 107 m 共350 ft兲 section. Since consolidation had taken place below the original embankment, but not on the edges where the expansion would be located, differential settlement between the two regions was a significant concern. Several options were considered as potential solutions, including a land bridge and realignment of the highway. Ultimately, a composite foundation system including a pilesupported section of the embankment was selected. Soft soils on the southeast side of the expansion were found to be shallower than those on the northeast side, and it was decided that a more conventional shallow foundation could be constructed. The soft soils were to be partially excavated and then loaded with a surcharge to accelerate consolidation. Following the surcharge, the southeast expansion would incorporate geofoam lightweight fill in order to reduce the load on the subsoils. The geofoam also extended across the entire base of the roadway to the edge of the column-supported section. The particular section of highway is bordered on the northwest by a small pond, and on the southeast by marshy terrain 共Fig. 1兲. A pile-supported LTP 0.9 m 共3 ft兲 thick was the foundation method selected for the westbound side of the highway, alongside the pond. The soil profile consisted of 9.1 m 共30 ft兲 of highly organic silt loams and peats underlain by about 6.1 m 共20 ft兲 of silty organic soils. Below that were 3.7 m 共12 ft兲 of loamy sand, underlain by about 10.7 m 共35 ft兲 of gravelly sand; a wellcemented sandstone was located 30.5 m 共100 ft兲 below the surface.

Instrumentation The area of the pile-supported embankment chosen for instrumentation was in the southwest end, between stations 168+ 00 and 168+ 50, because it was a good distance from potential vibrations caused by a railroad that could have affected measurements, and also because the backfill was thickest in that section, approximately 3.0 m 共10 ft兲, providing the largest load such that relative error would be minimized. In cross section view, the instrumented area was located beneath the shoulder/highway section of the embankment, rather than beneath the reinforced slope. Instrumentation was divided between two nearby “unit cells,” each defined by a group of six piles, Unit Cell 1 and Unit Cell 2

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / JANUARY 2010 / 27


only values up to November 30 are reported to eliminate effects of frozen soil.

unit cell 2

P4 BP_EPC

8 7 0.6 m

TP_EPC TP4

Foil strain gage

6 Concrete strain gage

5

2.1 m

TP3

P3

N

P2 BP_EPC 4 3 TP_EPC

TP2 Array of 10 strain gages geogrid layer GL3 2 Vibrating wire strain gage 1 Concrete strain gage P1

TP1

unit cell 1

Fig. 2. Plan view of the instrumented region

共Fig. 2兲, with a center-to-center pile spacing of 2.1 m 共7 ft兲 and a pile cap diameter of 0.6 m 共2 ft兲. In general, the separate unit cell design was chosen to achieve some redundancy of measurements. The sensor locations relative to the piles forming the unit cells were largely identical between the unit cells, allowing a comparison of measurements. Additionally, they served as a means to organize sensors with different sensing elements. For example, vibrating-wire EPCs were used in Unit 1, while semiconductorbased EPCs were used in Unit 2. A total of 48 sensors were installed throughout the LTP and on the supporting piles, including 12 EPCs, 6 settlement systems, 20 strain gauges on the geogrid, 2 strain gauges embedded in the concrete pipe piles, and 8 strain gauges on the pile walls. Eight EPCs, four in each unit cell, were installed at the base of the LTP at the same elevation as the pile cap, prior to the placement of a separator fabric. In each unit cell, two EPCs were located at the midpoint between two piles and two were located at the centroid of triangular pile group 共Fig. 2兲. Four EPCs were installed at the top of the platform, two in each unit cell: one above a pile, and another at the centroid 共Fig. 2兲. The data collection system consisted of one datalogger 共Campbell Scientific CR10X兲 and five multiplexers 共Campbell Scientific AM16/32兲, one vibrating wire interface 共Campbell Scientific AVW1兲, and one 16 MB storage module 共Campbell Scientific SM16M-ST兲. The datalogger was powered by a 12 V marinestyle battery, which was recharged by a regulated solar panel 共Campbell Scientific SP-2R兲 installed on top of the enclosure. A number of sensors were damaged during the construction process. All four foil strain gauges installed on the piles were not consistently operational; initial readings were obtained on two of the gauges, but further data were sporadic. One of the four vibrating wire strain gauges on the pile wall was damaged, as was one of the two embedment gauges. Of the 20 strain gauges that were installed on the geogrid, two did not survive the initial installation, and another 10, mostly on Geogrid Layer 1, did not survive the backfilling process. Data from one EPC 共BP_EPC 2兲 were not consistent. All readings began following the construction of the LTP on September 25; the loading from the LTP was not included. Data were collected for approximately nine months, but

Strain Gauges Four piles, labeled P1–P4 共Fig. 2兲, were instrumented with two axially oriented strain gauges on the outer pile wall, 305 mm 共12 in.兲 below the top of the pile. The instrumented piles were in the same row in the pile grid, parallel to the highway centerline. Three types of strain gauges were used in or on the piles. Two concrete embedment strain gauges 共Geokon 4200A-2兲 labeled P2C and P4C were installed, one in each of two piles, P2 and P4. The gauge locations were diametrically opposed in a line parallel to the highway centerline. Two types of strain gauges were installed on the pile walls; piles P1 and P2 were instrumented with the Geokon 4100A-2 vibrating wire strain gauge, while piles P3 and P4 were instrumented with Micromeasurements EA-06250BF-350 共option LE兲 resistive strain gauges. Vibrating wire gauges are quite rugged and survivability is high. A disadvantage is the reduced accuracy compared to foil gauges. For example, the sensitivity is typically given in a batch calibration factor, in which a small sample of a production batch is calibrated, and the average sensitivity of that sample is used for the entire batch. Previous work in which laboratory calibrations were performed on similar gauges has shown that the sensitivity can vary by more than 20% of the batch sensitivity. Unfortunately, individual calibration could not be performed because of time constraints. Prior to the installation of the gauges, dirt and corrosion were removed from the surrounding area on the pile with a grinder. For the foil gauges, a cyanoacrylate adhesive 共M-200兲 was used to attach the gauges to the pile walls. A waterproofing kit 共M-Coat F兲 provided butyl and neoprene rubber sheets, as well as a foil tape and a liquid air-drying nitrile rubber coating to seal the gauge and wiring from moisture. A spot-welder was used to attach the vibrating wire gauges. The mounting tabs were spot-welded. Immediately afterwards a layer of cyanoacrylate adhesive was applied to the surface of the tabs for corrosion protection. The coil assembly, which rests directly on top of the strain gauge, was held in place with mounting strips that were also spot welded directly to the pile wall. A caulk designed for use with steel was applied to the edges of the assembly before and after the strips were spot-welded in order to seal the interior of the gauge assembly. To protect all of the pile gauges from being accidentally detached, the cable was wrapped around the pile and secured to the wall in order to provide a frictional resisting force to accidental pulling. Finally, all cables were passed through a flexible conduit that led to the data collection box. A roll of the Synteen SF 80 geogrid was obtained from the field site and brought back to the laboratory for the installation of strain gauges. Geogrid Layers 1 and 3 were instrumented with resistive strain gauges 共EP-08-500GC-350兲. Synteen SF 80, which has an ultimate strength of 108 kN/m 共7,400 lb/ft兲 in the longitudinal direction, is composed of multifilament polyester yarns that are coated with a PVC material. An array of 10 gauges at five locations was oriented parallel to the highway centerline, centered over pile P2. Two gauges were installed at each location, one oriented parallel to the highway 共labeled L for longitudinal兲 and the other oriented perpendicular 共labeled T for transverse兲. The idea of this layout was to get a profile of the strain from the pile to the pile group centroid, and determine where the greatest strain was occurring. Centering the array over P2 allowed for some redundancy of measurements. The method used to install



The EPCs used in this project were of the hydraulic variety. Two thin circular plates are welded together at the periphery leaving a narrow disc-shaped void that is filled with fluid and connected to a pressure transducer through a short stem. The EPC contains a relatively incompressible fluid. Pressure develops in the fluid in response to the loading on the outer surface of the plates. The pressure measured by the transducer is assumed to approximate the average normal stress on the face of the cell. Because the stresses being measured in the soil were expected to show significant spatial variation, a sensor with a relatively small face was desired. Therefore, EPCs with a diameter of 117 mm 共4.6 in.兲, rather than the more common 229 mm 共9 in.兲 face, were chosen. Half of the EPCs used vibrating wire pressure transducers 共Geokon model # 4800-1X-170KPA兲, while the other half contained semiconductor-based pressure transducers 共Geokon model # 3500-1X-160KPA兲. Although factory sensitivities were supplied with the EPCs, the sensors were individually calibrated for soil conditions 共Labuz and Theroux 2005兲. At the bottom of the platform, EPCs were installed in four locations in each unit cell: two located at the centroids of the surrounding triangular pile group and the remaining two located near the midpoint between two adjacent piles. Care was taken in the instrumentation plan to ensure that there was sufficient distance between the sensors to minimize local stress effects from interactions. Again, the locations were duplicated in each unit cell so as to provide measurement redundancy. Vibrating-wire EPCs were used in Unit Cell 1, while semiconductor-based EPCs were used in Unit Cell 2. Both types of sensors were of identical geometry and nearly identical capacities, about 170 kPa 共25 psi兲 and 160 kPa 共23 psi兲, respectively. The installation of the EPCs involved several steps. First, a small hole was excavated. Several inches of sand identical to that used in the EPC soil calibration were placed and compacted in the same manner as in the calibration. The sensor was installed and leveled, then covered with an additional sand layer, which was also compacted. Initial readings were taken on all sensors. Cables were run through a flexible conduit to the data collection box.

Field Measurements Axial Strain and Load on Piles Data from the functioning gauges on the pile wall 共P1E, P1W, and P2E兲 followed a nearly identical trend, but they exhibited varying magnitudes. The backfilling process was observed by a field inspector, and Fig. 3 shows a plot of the pile gauge readings, including P2C, with time, and embankment height is included 共recall that no readings were taken as the LTP was constructed兲. The pile gauges measured different values of strain, and at the end of backfilling, the difference was especially significant for readings on Pile 1: strain in P1E 共495⫻ 10−6兲 was 60% greater than that of P1W 共303⫻ 10−6兲. The variation from an individual pile could be due to the variation in the reported sensitivity of a vibrating wire gauge, bending, or local effects from the pile cap.

500

P1E

fill

4

)

-6

P2E 400 3 P1W

300

2

P2C 200

Embankment (m)

Earth Pressure Cells

5

600

Δ strain (x10

strain gauges on the geogrid was adapted from a study of various installation methods that measured gauge survivability 共Warren et al. 2005兲. A silicone adhesive 共Dow Corning 3145兲 was used as the glue and waterproofing for the gauges.

1 100

0 25 Sep

0 11 Oct

27 Oct

12 Nov

28 Nov

14 Dec

Date

Fig. 3. Strain measurements on piles during backfilling

Consistent data were obtained from the two piles, where the mean axial strain of P1 共399⫻ 10−6兲 was within 10% of the strain in P2 共442⫻ 10−6兲. Assuming uniaxial stress and known properties of the steel 共E = 200 GPa兲 pile, and ignoring the data from the concrete embedment gauge, the axial load was estimated to be 475 kN 共107 kips兲 in Pile 1 and 526 kN 共118 kips兲 in Pile 2, which exceeded the design load of 400 kN 共90 kips兲 based on the tributary area. One explanation is again the variation of sensitivity for vibrating wire strain gauges. Another explanation of the large estimated load is that the piles from which the measurements were taken were supporting a load from outside their tributary area, the hexagonal area surrounding each pile from which the 400 kN was calculated. Because of the three-dimensional 共3D兲 aspect of the pile layout, piles in the row that were instrumented could support a greater load than piles in other rows. Nevertheless, it appeared that the load transfer from the embankment to the piles was complete. Earth Pressure at Base of LTP Earth pressure was measured at the base of the platform between two adjacent piles and at the centroid of three piles, and at the top of the platform directly above the piles and at the centroid. It should be noted that all EPC readings were corrected for temperature. Fig. 2 shows the locations of the vibrating wire EPCs 共1, 3, 4兲 and the semiconductor EPCs 共5, 6, 7, 8兲 at the base of the LTP in Unit Cells 1 and 2, respectively. Half of the EPCs at the base of the platform were installed at the centroid of a pile group 共1, 4, 5, 8兲, while the other half 共3, 6, 7兲 were installed at the midpoint between two adjacent piles. Plots of EPC readings between piles at the base of the platform for Unit Cells 1 and 2 are shown in Fig. 4共a and b兲, with the estimated vertical stress due to overburden, ␥H. 共The data from the EPCs in Unit Cell 2, BP_EPC5, 6, 7, 8, were lost until October 25兲 The midpoint EPCs 共3, 6, 7兲 at the base of the LTP showed similar behavior. As indicated in Fig. 4共a兲, an increase in the readings was observed at the end of each period of fill placement, followed by a slight decline in the readings. After backfilling was completed on October 20, readings decreased steadily to near zero 关Fig. 4共a兲兴 or a value of 15 kPa 共300 psf兲 or less 关Fig. 4共b兲兴. For the centroid locations, the EPC in Unit Cell 1 共BP_EPC3兲 exhibited similar readings of the vertical stress near zero, while the readings in Unit Cell 2 from BP_EPC6 and BP_EPC7 were about 20 kPa 共400 psf兲. Vertical stress readings at the base of the LTP correspond



-0.05

fill

-0.04

40

4

)

∆ strain (x10

-6

fill

60

Δσv (kPa)

5

-0.03 3

GL3_1L -0.02

GL3_3L 2

-0.01

BP_EPC1 GL3_3T

BP

20

EPC4

1

0 GL3_5L

BP_EPC3 0.01 25 Sep

0

(a)

25 Sep

11 Oct

27 Oct

Embankment (m)

80

12 Nov

28 Nov

0 11 Oct

27 Oct

12 Nov

28 Nov

14 Dec

Date

14 Dec

Date

Fig. 6. Strain in Geogrid Layer 3. The minus sign indicates tension. 80

fill

Earth Pressure at Top of LTP and Strain in Geogrid

Δσv (kPa)

60

40

BP_EPC6 BP_EPC7

20

BP_EPC5

BP_EPC8 0

(b)

25 Sep

11 Oct

27 Oct

12 Nov

28 Nov

14 Dec

Date

Fig. 4. Vertical stress between piles at the base of the platform: 共a兲 Unit Cell 1; 共b兲 Unit Cell 2

with the expected design: stress develops above the EPCs initially until enough embankment fill is available to transfer the increasing embankment load to the stiffer regions of the platform surrounding the piles. As the soil settles, the regions of fill involved in arching stiffen in response to the load, relieving the stress between the piles. The centroid EPCs in Unit Cell 2 registered some vertical stress, but the other EPCs measured near zero. Nearly all the embankment load appeared to be transferred to the pile.

Recall that four EPCs were installed at the top of the platform, two in each unit cell; one EPC per unit cell was installed above a pile, and the other was located at the centroid of the triangular pile group. The two EPCs at the centroid location showed behavior very similar to that of the midpoint EPCs at the base of the LTP: vertical stress at the centroid of the piles was essentially zero 共Fig. 5兲. The EPCs above the piles at the top of the LTP behaved differently. Fig. 5 is plotted with the estimated vertical stress due to overburden 共␥H兲; it is clear that the readings increased substantially during backfilling and reached a plateau of between 240 kPa 共5,040 psf兲 and 450 kPa 共9,360 psf兲, 4–7.5 times the vertical stress due to the backfill. Apparently, the action of the stiff pile acted through the thickness of the platform and the vertical stress was drawn to this region, with virtually no vertical stress present at the centroid of the pile group. Thus, the platform was not acting as a beam but as a 3D stress diffuser 共Villaggio 1981兲. The design of the LTP should be concerned with bearing capacity from the significant load intensity on the pile cap. Strain measured on Geogrid Layer 3 共GL3兲 was found to be generally in the range of 1–2%, well below the design strain of 5% 共Fig. 6兲. The strains in the longitudinal direction were generally larger than those in the transverse direction. Most of the strain occurred during backfilling. Readings did not appear to depend on the location of the gauge. No gauges survived on Geogrid Layer 1 共GL1兲. Although the geogrid provided minimal vertical support to the system, it does serve the critical purpose of added confinement to the highly stressed material in the platform.

500

TP_EPC1

Numerical Modeling

Δσv (kPa)

400

300

TP_EPC3 200

100

TP_EPC2

fill

TP_EPC4 0 25 Sep

11 Oct

27 Oct

12 Nov

28 Nov

14 Dec

Date

Fig. 5. Vertical stress above 共TP_EPC1,3兲 and between 共TP_EPC2,4兲 piles at the top of the platform in Unit Cells 1 and 2

FE simulations using elastic and elastoplastic constitutive models were performed with STRAND7, Standard Version 2.3, a commercially available FE software package. Quasi-static 3D analyses were performed with eight-node isoparametric brick elements for the platform and the embankment, and four-node plate elements of negligible bending stiffness for the grid. The FE mesh was refined in order to have accurate numerical results, verifying that the displacement and stress fields were significantly regular without any substantial gradients. The FE mesh was composed of 38,719 nodes, 3,780 plate elements, and 37,800 bricks elements, and this region captured the 3D aspects of the structure. The imposed boundary conditions were



Table 1. Mechanical Properties of the System Young’s modulus

Gravel LTP Sand backfill Grid

共MPa兲

共psi兲

Poisson’s ratio

100 20 0.14

14,500 2,900 20

0.3 0.3 0.3

Unit weight 共kN/ m3兲

共pcf兲

20 20 1.9

130 130 12

1.

The vertical plane at the right boundary of the embankment was constrained by rollers that prevented horizontal displacements; and 2. The platform was supported by piles with elastic deformability. No constraints were placed on the slope and on the top of the embankment. The first numerical analysis considered the materials to be isotropic, defined by the engineering constants shown in Table 1. The properties of the granular materials 共the gravel forming the LTP and the sand backfill兲 were measured in triaxial compression experiments with bender/extender element testing. The properties of the grid were estimated from the manufacturer’s specifications. A second analysis with a coarser grid considered the elastoplastic behavior of the granular materials through a Mohr-Coulomb failure criterion with zero dilatancy. However, the response of the system remained essentially unchanged from that of the elastic analysis 共neglecting numerical instabilities兲, because material failure was confined to small regions surrounding the pile caps.

Transversely Isotropic LTP To account for the construction process and to more closely match the field observations 共nonuniform pile loads, large vertical stress above the piles, and negligible vertical stress between the piles兲, a third numerical analysis considered the LTP composed of a transversely isotropic material: the five elastic constants are the Young’s modulus E p and Poisson’s ratio ␯ p in the x-z symmetry plane; the Young’s modulus Ey and Poisson’s ratio ␯y in the y-direction; and the shear modulus Gzp in the z-direction. It was assumed that the Young’s modulus and Poisson’s ratio in the transverse direction 共E p = 10 MPa or 1,450 psi and ␯ p = 0.03兲 was 10% of the measured values 共Ey = 100 MPa or 14,500 psi and ␯y = 0.3兲, and Gzp = 41 MPa 共6,000 psi兲. The support provided by the piles was modeled as springs, with stiffness equal to the overall axial stiffness of the steel pile filled with concrete, assuming a modular ratio n = Es / Ec = 15. Only the numerical results from the transversely isotropic platform will be presented because the numerical predictions were closer to the field data, although the other two models provided similar qualitative results. Due to the very high relative stiffness of the piles with respect to the surrounding ground, the vertical displacements at the top of the piles were negligible, enforcing the fundamental role of the piles in reducing the deformability of the embankment. The nodes at the bottom of platform between the piles exhibited a vertical displacement of only 13 mm 共0.5 in.兲, about 10 times less than the values indicated by the settlement systems 共Wachman and Labuz 2008兲. Because consolidation of the soft organic soils was not modeled, a consequence of the limited settlement from the numerical simulation will be reduced strains in the grid. Nevertheless, enhanced displacement gradients were noted close to the pile caps, suggesting critical situations in terms of strain 共and stress兲 concentrations. As suggested by Hewlett and Randolph 共1988兲,

217 kN

120 kN

341 kN

307 kN 298 kN

363 kN

294 kN

Fig. 7. Axial loads on piles from the numerical analysis

bearing capacity at the pile cap is a critical issue for design. Axial loads on the piles are a basic comparison that can be made between the field data and the numerical modeling. It should be emphasized that the geometry and the three dimensionality of the structure produces a variation in pile loads 共Fig. 7兲. The maximum load on a pile was 363 kN 共81.5 kips兲, about 25% lower than the experimental result. Several factors could explain this result. A variation of pile stiffness could account for one pile attracting load from its neighbors. In addition, a 20% variation in vibrating wire gauge sensitivity is within this range. In any case, the three dimensionality of the problem is demonstrated, and the pile loads are not simply the tributary area multiplied by the overburden stress. The vertical stress distributions through the height of the platform along four different vertical positions are presented in Fig. 8. Three of the distributions 共a, b, c兲 are above the pile and one 共d兲 is between two piles where the midpoint EPCs were positioned. Critical regions with very high stress concentrations are located at the boundary of the pile caps. A significant volume of platform among the pile caps is almost stress-free and appears necessary only as filling material that provides confinement. The results between the modeling and the field measurements qualitatively agree. At the elevation corresponding to the bottom of the platform 共the top of the pile兲, the vertical stress distribution varies as a stiff footing on an elastic soil; the stress at the edge is three times the stress at the center. In addition, the vertical stress between the piles is almost zero. The vertical stress distribution at the top of the platform is shown in Fig. 9. The field measurements indicated a vertical stress above the piles of 30–50 times larger than between the piles at the top of the platform, and similar behavior was captured by the numerical simulation. The platform area above the pile acts as an extension of the pile, and the stress trajectories tend to spread to develop the arching effect above the platform. The effect of the stiff pile is evident in the distribution of vertical stress.

Structural Behavior Arching within the system was pronounced at the top of platform over the pile caps, where the vertical stress was measured at the top layer of the platform, and thus arching was achieved mainly in the embankment. Both experimental 共Fig. 5兲 and numerical



500

400

σyy (kPa)

300

200

100

0 0

100

200

300

400

500

600

700

-100

(a)

Distance (cm)

Distance from the platform base (cm)

100

80

d

(b)

60

Fig. 9. 共a兲 Vertical stress across the width of LTP; 共b兲 section view of LTP where vertical stress was evaluated

40

c

20

a

b

0 -0.2

0

0.2

0.4

0.6

0.8

1

1.2

Normal stress yy (MPa)

Fig. 8. Vertical stress distributions at four positions within the LTP. Positions a, b, and c are above the pile cap, and position d is between two piles.

共Fig. 9兲 results showed that the vertical stresses were negligible in the upper layer of the platform, excluding the regions associated with the pile extensions. Due to the significant axial stiffness of the piles in relation to the surrounding soil, it appeared that the pile extended through the platform and 3D soil arching 共vaulting兲 developed as a result of differential settlements between the stiff pile extensions and the soil. The mechanism of transferring the load is shown schematically in Fig. 10共a兲, which depicts a “load-carrying skeleton.” This sketch was produced from the numerical model, which indicated that the vertical stress was almost zero up to a distance of 0.5 m 共1.7 ft兲 above the top of the platform within regions not directly over the piles. The arching effect identified “vaults” in the embankment with a very low 共0.5 m兲 rise, and not semicircular as sometimes assumed. As shown in Fig. 10共b兲, the vertical stress within the embankment was affected by the pile extension, even at Level 1, which is at a distance of approximately 0.6 m 共2 ft兲 above the LTP. However, the vertical stress at Level 1 coincided with the direction of the major principal stress; this was not the case below Level 1. At Level 2, a distance of 1.5 m 共5 ft兲 above the LTP, the perturbation of the vertical stress due to the pile extension was diminished. Finally, at Level 3, a distance of 2.4 m 共8 ft兲 above the platform, the stress field was regular. The action of the LTP is critical above the piles but not between them, as this part of the platform is lightly stressed. Thus, it does not act as a beam but as part of a vaulted arch, and the problem of stress diffusion becomes dominant 共Villaggio 1981兲. The highly stressed region above the pile will require sufficient strength, and the confinement provided by the geogrid should be

useful. The volume of the platform between the ideal pile extensions simply works as filling that provides confinement, which is important for the LTP performance. The vaulted arch extends into the embankment, although it appears to be shallow based on the numerical simulations. This shallow arch means that lateral confinement is critical, and in the interior of the platform the thrust from the vaulted arch is self-equilibrated if uniformly loaded. However, if the geometry is nonsymmetric, or if the embankment height is not sufficient for uniform diffusion of the load, then a self-equilibrated system is not formed, which is the case near the boundaries. The reinforcement needs to provide sufficient horizontal force 共confinement兲. A membrane effect of the geosynthetic reinforcement was clearly evident from the numerical results, but according to the simulation, the design of the grid was more than sufficient. Indeed, the stresses and strains in the reinforcement were negligible within the model, especially those located at the lowest position in the platform. However, the numerical model did not capture the consolidation behavior of the very soft organic soil, and the actual vertical displacements may have been an order of magnitude larger than the values from the numerical simulation. Thus, the numerical results gave maximum tensile strains of 0.1%, about one-tenth of the strains experimentally observed. Nonetheless, the vertical load carried by the geogrid was small, but the grid in the bottom part of the embankment could reduce problems at the base of slope and make a significant contribution at the end of the platform. It should be noted that this column-supported embankment is a 3D problem, and erroneous conclusions could be reached from the 2D case. For example, the classical mechanism of arching due to Janssen 共1895兲, which is also attributed to Marston and Anderson 共1913兲, cannot be applied to the 3D case because of neighboring unit cells. Furthermore, it is apparent that the problem is dictated by stress diffusion, which was detected experimentally with the EPCs and confirmed numerically. Even though the linear elastic model cannot accurately describe the stress states around the pile caps because failure is not considered, the simplified system captured remarkably well the structural behavior suggested by the field data.



column, the grid increases the strength in the vertical direction due to lateral confinement.

level 3

Acknowledgments level 2

Partial support was provided by the Minnesota Department of Transportation 共MnDOT兲. Special thanks are extended to the Technical Advisory Panel, especially Richard Lamb, Derrick Dasenbrock, Gary Person, and Glenn Engstrom. This paper represents the results of research conducted by the writers and does not necessarily represent the views of MnDOT. Furthermore, products mentioned in the paper are stated for completeness only, and not as an endorsement.

level 1 soil arching

0.5 m

(a)

References

300

level 1

250

level 2 level 3

σyy (kPa)

200

150

100

50

0 0

100

(b)

200

300

400

500

600

700

Distance (cm)

Fig. 10. 共a兲 Schematic representation of the load transfer mechanism; 共b兲 vertical stress in the fill at various levels

Conclusions The piles-platform-embankment system must be designed to provide stability for its intended function 共e.g., roadway兲 under the self-weight and service load. To fit the requirements, a “loadcarrying skeleton” is formed within the overall structure and this attracts stresses in the material above the piles and the loads to the piles. Within the platform, a region of highly stressed material forms such that a pile extension appears, where vault arching arises between the piles. The geogrid within the platform performs two important functions: 共1兲 it provides confinement within the LTP for pile extension, and 共2兲 it generates horizontal forces needed to form the vaulted arch. Of course, the horizontal forces are self-equilibrated for adjacent vaults if they are loaded uniformly, but if the geometry is nonsymmetric, or if the embankment height is not sufficient for uniform diffusion of the load, then a self-equilibrated system is not formed, which is the case near the boundaries 共e.g., the edge of the platform兲. The platform is important mainly because of the high vertical stress from the loading transferred by the vaultlike arch, and this material behaves as an extension of the piles in the form of highly stressed columns. The other material in the platform, not above the piles, is almost stress-free, and therefore, this material acts simply as a filler, although it does impart confinement to the highly stressed regions. Although the geogrid provides minimal vertical support to the system, it does serve the critical purpose of added confinement to the highly stressed material in the platform. Similar to stirrups in a reinforced concrete

Collin, J. G. 共2004兲. “Column supported embankment design considerations.” Proc., 52nd Annual Geotechnical Engineering Conf., J. F. Labuz and J. G. Bentler, eds., Univ. of Minnesota, Minneapolis, 51– 78. Fluet, J. E., Christopher, B. R., and Slaters, A. R. 共1986兲. “Geosynthetic stress-strain response under embankment loading conditions.” Proc., 3rd Int. Conf. on Geotextiles, Geomembranes, and Related Products, Vienna, Austria, Balkema,Rotterdam, The Netherlands, 175–180. Gartung, E., Verspohl, J., and Alexiew, D. 共1996兲. “Performance of a geogrid reinforced railway embankment of piles.” Proc., 14th Int. Conf. on Soil Mechanics and Foundation Engineering, Vol. 3, Swets & Zeitlinger, Hamburg, Germany, 1731–1736. Han, J., and Gabr, M. A. 共2002兲. “Numerical analysis of geosyntheticreinforced and pile-supported earth platforms over soft soil.” J. Geotech. Geoenviron. Eng., 128共1兲, 44–53. Hewlett, W. J., and Randolph, M. F. 共1988兲. “Analysis of piled embankments.” Ground Eng., 21共3兲, 12–18. Janssen, H. A. 共1895兲. “Versuche über getreidedruck in silozellen.” Z. Ver. Dtsch. Ing., 39, 1045–1049. Jones, C. J. F. P., Lawson, C. R., and Ayres, D. J. 共1990兲. “Geotextile reinforced piled embankments.” Proc., 4th Int. Conf. on Geotextiles, Geomembranes, and Related Products, Balkema, Rotterdam, The Netherlands, 155–160. Labuz, J. F., and Theroux, B. 共2005兲. “Laboratory calibration of earth pressure cells.” Geotech. Test. J., 28共2兲, 188–196. Love, J., and Milligan, G. 共2003兲. “Design methods for basally reinforced pile-supported embankments over soft ground.” Ground Eng., 36共3兲, 39–43. Low, B. K., Tang, S. K., and Choa, V. 共1994兲. “Arching in piled embankments.” J. Geotech. Engrg., 120共11兲, 1917–1938. Maddison, J. D., Jones, D. B., Bell, A. L., and Jenner, C. G. 共1996兲. “Design and performance of an embankment supported using low strength geogrids and vibro concrete columns.” Proc., 1st European Geosynthetics Conf., Maastricht, The Netherlands, Balkema, Rotterdam, The Netherlands, 325–332. Marston, A., and Anderson, A. O. 共1913兲. The theory of loads on pipes in ditches and test of cement and clay drain tile and sewer pipe, bulletin 31, Iowa Engineering Experiment Station, Ames, Iowa. Rathmayer, H. 共1975兲. “Piled embankment supported by single pile caps.” Proc., Conf. on Soil Mechanics and Foundation Engineering, Vol. 1, Istanbul, Turkey, 283–290. Reid, W. M., and Buchanan, N. W. 共1984兲. “Bridge approach support piling.” Piling and ground treatment, Thomas Telford, London, 267– 274. Russell, D., and Pierpoint, N. 共1997兲. “An assessment of design methods for piled embankments.” Ground Eng., 30共11兲, 39–44. Stewart, M. E., Navin, M. P., and Filz, G. M. 共2004兲. “Analysis of a



column-supported test embankment at the I-95/Route 1 interchange.” Proc., Geo-Trans 2004, Los Angeles, ASCE, Reston, Va., 1337–1364. van Eekelen, S. J. M., Bezuijen, A., and Alexiew, D. 共2008兲. “Piled embankments in The Netherlands, a full-scale test, comparing 2 years of measurements with design calculations.” Proc., EuroGeo4, Edinburgh, UK Chapter, International Geosynthetics Society. van Eekelen, S. J. M., Bezuijen, A., and Oung, O. 共2003兲. “Arching in piled embankments: Experiments and design calculations.” Proc., Foundations: Innovations, Observations, Design and Practice, Dundee, Scotland, Thomas Telford, London, 885–894.

Villaggio, P. 共1981兲. “Stress diffusion in masonry walls.” J. Struct. Mech., 9共4兲, 439–450. Wachman, G., and Labuz, J. F. 共2008兲. “TH 241 column-supported embankment.” Minnesota Department of Transportation Rep. No. CTS #08-11, 具http://www.cts.umn.edu/Publications/ResearchReports/ reportdetail.html?id⫽1620典共Nov. 2, 2008兲. Warren, K. A., Brooks, J. A., and Howard, I. L. 共2005兲. “Survivability of foil strain gages mounted on geosynthetics under full-scale construction loads.” Proc., Geo-Frontiers Conf. 2005, Austin, Tex., ASCE, Reston, Va., 4085–4090.

