Lessons from a failure case of an excavated floodway supported by precast cantilever pile walls

Engineering Geology 209 (2016) 106–118

Contents lists available at ScienceDirect

Engineering Geology journal homepage: www.elsevier.com/locate/enggeo

Lessons from a failure case of an excavated floodway supported by precast cantilever pile walls Meng-Hsiung Cheng a,b, Ming-Wan Huang c, Yii-Wen Pan b,⁎, Jyh-Jong Liao b a b c

Water Resources Planning Institute, Water Resources Agency, 1340 Jhong-jheng Road, Wu-fong, Taichung 413, Taiwan Department of Civil Engineering, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan Disaster Prevention & Water Environment Research Center, National Chiao Tung University, 1001 University Road, Hsinchu 300, Taiwan

a r t i c l e

i n f o

Article history: Received 17 March 2016 Received in revised form 27 May 2016 Accepted 27 May 2016 Available online 29 May 2016 Keywords: Floodway Precast cantilever pile walls Failure mechanism Slope stability

a b s t r a c t The Gangweigou floodway, across the Quiren and Rende districts in Tainan, Taiwan, was designed to bypass partial discharge in the Gangweigou River during heavy rainfall events for the purpose of flood control. This floodway was an excavated open channel supported by single or double rows of precast RC cantilever pile walls on each side of its banks; these piles were installed by the jetting-assisted method. Twenty days after the completion of the floodway, it was put to work for the first time during a heavy rainfall. Unexpectedly, the cantilever pile walls in several sections failed and caused the collapse of the slope behind the walls. This paper explores the failure mechanism and the real causes of failure in this case. Based on the conditions of hydrology and hydraulics, failure processes and the results of stability analyses, it was concluded that ignoring a possible bed incision or bank erosion was likely the primary problem with the improper design. The most critical condition was when the water level in the floodway quickly dropped, which resulted in rapid drawdown. The failure of the case could have been avoided if both the rapid drawdown condition and the potential of the channel bed incision were considered. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Cantilever pile walls often serve as retaining structures for various civil engineering or hydraulic engineering purposes. The use of precast reinforced concrete (RC) piles interconnected with adjacent units is a feasible technique for the formation of a cantilever pile wall. The tops of these jointed wall units can then be connected with a continuous RC girder as a pile cap. Compared with cast-in-situ piles, precast RC pile walls may have several advantages, including better quality control, shorter construction time and lower cost. To work as a flood protection structure, the use of installed walls can avoid large amounts of excavation. Therefore, the use of precast RC walls is a good choice for flood protection when the groundwater table is high and when the available land is limited. The Gangweigou River is one of the main tributaries of the Erren River in southern Taiwan (Fig. 1). To mitigate repeated flooding in the midstream region of the Gangweigou River, the River Management Office commenced a flood control plan by building the Gangweigou floodway to bypass partial discharge exceeding the flood design in the upstream Gangweigou River directly into the Erren River. ⁎ Corresponding author. E-mail addresses: [email protected] (M.-H. Cheng), [email protected] (M.-W. Huang), [email protected] (Y.-W. Pan), [email protected] (J.-J. Liao).

http://dx.doi.org/10.1016/j.enggeo.2016.05.014 0013-7952/© 2016 Elsevier B.V. All rights reserved.

The Gangweigou floodway was constructed by excavating the existing canal; the earth material is mainly composed of silty sand with clay and silty clay with sand. The channel banks were supported by one to two parallel series of precast RC cantilever pile walls on both sides (Fig. 2). The geometry of the floodway was an open channel with a compound rectangular section. The precast RC cantilever pile walls were used as the retaining structure for flood protection in the Gangweigou floodway. The precast RC piles were installed by the jetting-assisted method to drive them to pass through the soil-containing cobbles occasionally. Very soon after the completion of the floodway, a heavy rainfall event occurred, and the floodway was put to work for the first time. Unexpectedly, the cantilever pile walls in a few sections failed and caused the collapse of the slope behind the walls. This paper attempts to explore the failure mechanism and the real causes of failure in this case. The understanding of the failure causes in this case study may help improve the design of flood protection works using cantilever pile walls in the future. 2. Failure modes of the cantilever pile walls and methods for analysis Sheet pile walls can be either cantilever pile walls or anchored pile walls, depending on whether the anchors are used. Usually, a cantilever pile wall is driven into the soil to provide lateral support below the excavation bottom; the wall has to resist both overturning and sliding. In

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

107

Fig. 1. Maps of the location.

addition, the material stresses in the wall must be within acceptable values. Among possible applications, the functions of the cantilever pile walls may include the following: 1) to maintain the stability for deep excavation as a retaining wall; 2) to reduce underground flow as a seepage barrier wall; 3) to replace a levee for flood protection; and 4) to serve as a bulkhead wall in a waterfront (Dutta and Vaidya, 2003).

When a cantilever pile wall system is used in a waterfront, either along riverbanks or in a harbor, the cantilever pile wall must maintain stability and prevent the failure and excessive settlement of the ground behind the wall. The loading on the wall should consider the lateral earth pressure, the water pressure acting on the wall, and the surcharge on the ground surface. In general, the penetration depth of the

Fig. 2. Plane view and typical cross sections of the site.

108

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

cantilever piles should be deep enough to ensure stability of the structure. The cantilever piles can be made of steel or reinforced concrete. The failure modes of cantilever pile walls as supportive structures for manmade channels or waterfront banks may include 1) global sliding failure, 2) cap or tie-back rod breaks, 3) wall rotation or toe and berm failure, 4) seepage, and 5) joint separation or waterline failure (CMIWAC, 2014; USACE, 1994). The failure of a cantilever pile wall can also be a result of overstress because the bending moment in the cantilever is too large. For the failure case of the flood protection structure in the Gangweigou floodway, the deterioration of the wall material is unlikely to be the reason for the occurrence of the failure incident. The major reasons for the failure case were suspected to be the overstressed wall material or the insufficient penetration depth that arose from global sliding, unbalanced earth pressures or structural failure. When the penetration depth of the wall is shallower than the sliding plane, the retaining wall has no effect in preventing sliding failure. In that case, the wall structure would be displaced along with the sliding mass. This type of failure is referred to as global sliding failure. Hence, the wall penetration depth might affect the failure mechanism of the retaining wall. Using centrifuge models, Bolton and Powrie (1987) investigated how the penetration depth of a cantilever retaining wall in a clayey soil would affect the failure mode of the retaining system during the excavation stage. For a shallow penetration depth, the rotation angle of the retaining wall would be large and could cause the separation of the wall and the soil behind; as a result, tensile cracks could be formed. If the tensile cracks were filled with water, the retaining structure could collapse quickly. If the wall penetration depth were deep enough, there would be no tensile cracks formed because the wall would maintain contact with the soil behind it. Along with expansive deformation, the soil could develop a negative pore water pressure and remain temporarily stable. As surrounding water gradually enters the soil, the negative pore water pressure disappears; the retaining structure may lose its stability as a result. Liu (2002) also conducted centrifuge model tests to study the failure behavior of excavation supported by a cantilever wall. His results showed that the earth pressure does not increase in proportion with increasing wall deformation after the deformation becomes very large. No obvious heave was observed in his model tests. He observed that the starting point of the failure plane was approximately the midpoint of the penetration depth, which made a 60-degree angle with the horizontal direction and then extended all the way to the ground surface. The designs of cantilever sheet pile walls are usually based on the force and moment equilibrium by considering the active and passive earth pressures. The magnitude of the earth pressure depends on the strength properties of the soil, the interaction at the soil–structure

interface, and the level of the deformations in the soil–structure system. In reality, the in situ lateral pressure distribution σh can be complicated. The distribution of σh is usually simplified for the purpose of stability analysis. There are three types of methods for the analysis of cantilever pile walls: namely, the full method, the simplified method, and the gradual method (or the rectilinear net pressure method), classified according to the adopted distribution of σh (Alejo, 2013; Bowles, 1988; King, 1995; Padfield and Mair, 1984). The full method takes the rotation point (i.e., point R in Fig. 3a) of the sheet pile wall as the limit point. It assumes that σh within the ground surface and the rotation point is in the active state, and σh within the excavation bottom and the rotation point is in the passive state (as shown in Fig. 3a). The active and passive sides are reversed below the rotation point. The simplified method (as shown in Fig. 3b) is a simplified version of the full method; an equivalent concentrated force is applied to the bottom of the wall to consider the reversed σh distribution below the rotation point. The gradual method considers that the wall under the excavation bottom is simultaneously subjected to lateral pressures from both sides and assumes that the net pressure below the excavation bottom follows a rectilinear distribution. The common method used in the USA (named “USA method” by Day (1999)) belongs to the category of gradual methods; Fig. 3c shows the net pressure acting on the wall in the method. The depth of penetration can be obtained by estimating σh at the bottom of the wall and satisfying all equilibrium equations, (Bica and Clayton, 1989; Day, 1999; King, 1995). The designers of the Gangweigou floodway project adopted the simplified method for the design of the cantilever pile walls, neglecting the possible water pressure difference on both sides of the walls. In this Gangweigou floodway case, the jetting-assisted method was used for pile driving; this method helps prevent pile damage caused by the occasional cobbles in the soil. Compared with pile driving with a hammer, the water jet method is an effective means (time-saving, energy-saving, and less shaking) for driving concrete sheet piles into the strata of dense sand or hard clay (Tsinker, 1988). However, the construction standards for the jetting-assisted method are still under development (Shepley and Bolton, 2014). The in situ soil structure is likely significantly altered owing to water jetting: the zone immediately below the water jet may turn into a soil/water mixture and flow upward along the pile, lubricating the pile; soil liquefaction and seepage flow are expected to affect the original soil characteristics (Shepley and Bolton, 2014; Tsinker, 1988). Tsinker (1988) noted that the frictional resistance of a jetting-driven pile would be lower than that of an impact-driven pile. Some studies were undertaken to provide the installation results of jetting-driven piles (Bourdouxhe-Barnich et al., 2002) to establish

Fig. 3. Lateral earth-pressure distribution assumed in the major methods for the analysis of a cantilever pile walls: (a) Full method (redrawn from Day (1999)); (b) Simplified method (redrawn from Alejo (2013)); (c) Gradual method (redrawn from Day (1999)).

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

the mechanisms governing water jetting (Shepley and Bolton, 2014) and realize the lateral load behavior (p-y curves) of jetting-driven piles (Hameed et al., 2000). 3. Project background 3.1. Flood mitigation plan The region in the Gangweigou River has suffered from flooding from time to time, which has caused severe damage and inconvenience for the residents of the region. According to the planning report by the river management agency in charge, the reason for the flooding was that the high water level in the mainstream of the Erren River during a flood tends to block the flow from the Gangweigou River into the mainstream, thus resulting in flooding in the midstream of the Gangweigou River. A flood mitigation plan was proposed to build the Gangweigou floodway to bypass the excess discharge from the midstream of the Gangweigou River to the Erren River ahead of the junction of the two rivers (Fig. 1). In addition to the Gangweigou floodway, dredging in the upstream and a flood detention pool in the downstream of the Gangweigou River were the matching plans in the flood management project. The total budget of this project was approximately 0.92 billion NT dollars (approximately 30 million US dollars). The entire project was completed on July 27, 2014. Unexpectedly, the floodway collapsed during the first time it was put to work after its completion. The failure took place on August 13, 2014, just 20 days after the date of completion. The failure cause of the Gangweigou floodway is the main focus of this paper. 3.2. Site topography and geology The headwater of the Gangweigou River lies in the Guanmiao district of Tainan city, Taiwan. This river flows heading southwest and merges with the Erren River. The catchment area of the Gangweigou River, covering the Guanmiao, Quiren and Rende districts, is approximately 37 km2 with a total trunk length of 17 km. The region in the case study is within the Chianan Plain in southwestern Taiwan (Fig. 1). The Chianan Plain contains the deposition of the sediments eroded from the streams in the eastern terrace and extends all the way to the west coast. The entire region in the Gangweigou floodway is covered by Holocene alluvium, the Tainan formation, which is composed of gravel, sand and clay with a thickness from 16 to 30 m. According to C14 dating, the deposition age of the Tainan formation is 5800–7200 years (Chen, 1992). During the design stage, 10 borehole explorations with the Standard Penetration Test (SPT) were conducted; soils were sampled by a split spoon sampler and by 3-in thin tubes. The samples obtained by the split spoon sampler were used for the determination of index properties including water content, unit weight, grain-size distribution and

109

the Atterberg limits. The samples obtained by thin tubes were used for direct shear tests, consolidation tests, unconfined compression tests and unconsolidated undrained triaxial compression tests. The soil profile along the floodway is composed of clay with sand, silty sand with clay, and mudstone, in sequence as shown in Table 1. Four in situ variable head permeability tests were also carried out on silty sand layers inside bore holes. The field hydraulic conductivity of silty sand layers was within 2–9 × 10− 5 cm/s. The groundwater table was 2.3–6.3 m below the ground surface during the ground exploration (Water resource agency, 2010). 3.3. Hydrological analysis of the floodway The floodway is mainly an open channel with compound rectangular sections, except for a few sections that are culvert boxes to underpass facilities. The width of the channel cross section is 25 m. To control the water level to start the bypass discharge, a river barrage was built in the Gangweigou River at the location of the bypass floodway. When the discharge in the Gangweigou River exceeds 6 m3/s, the excess flood discharge would flow from an overflow inlet upstream of the river barrage, entering a lateral still basin first and then into the floodway connecting to the Erren River. The Gangweigou floodway was excavated, with a total length of 3769 m. The floodway project was divided into six construction sectors. The sixth sector (0K-135–0K+620) was the one in which the floodway would merge with the Erren River and was also where the failure took place (Fig. 2). The excavated bed slope of the floodway was 1/1800; the bed slope of the floodway merging into the Erren River (0K-135– 0K+000) maintained the original natural slope, 1/54. In the planning stage, it was assumed that (1) the rainfall intensity in the whole area is uniform, and (2) the river water level reaches 8.45 m for the design flood (with 10-year return period); this flood would submerge the intersection region with floodway from station 0K+000 to 0K+050. In the design stage, it was expected that the flow velocity would not be too high owing to the mild slope of the floodway and the submerged flow at the flow junction. Furthermore, the maximum flow velocity should not exceed 2.6 m/s and would not cause a significant erosion problem. Consequently, the floodway bed was not protected by any ground sill. 4. The precast cantilever walls of the floodway 4.1. Design of the precast cantilever walls The maximum excavation depth of the Gangweigou floodway exceeds 11 m. The precast cantilever wall system was adopted as the flood protection structure for the following reasons: 1) to reduce the excavation volume of the soil mass in the channel section; 2) to cut down the cost and construction time; and 3) to avoid pile damage by cobbles

Table 1 Parameters of the simplified soil profile. Grain size distribution

Soil description Silty clay with sand (CL2) Silty sand with clay (SM3) Silty sand with clay (SM4) Silty clay with sand (CL5) Mudstone

Atterberg limits

Depth form ground surface (m)

SPT N value

Bulk unit weight (kN/m3)

Saturated unit weight (kN/m3)

Sand ≧#200 (%)

Clay ˂#200 (%)

Liquid limit (%)

Plastic limit (%)

Plasticity Index

0–3.6

3–4

17.66

19.43

2–3

97–98

40–41

20–23

20–24

3.6–7.3

7–19

19.13

21.04

56–62

38–44

–

–

7.3–12.5

9–25

19.42

21.37

55–61

39–45

–

12.5–22.2

5–8

17.66

19.43

4–17

83–96

25–36

22.2–30.0

˃50

23.15

23.85

–

–

Cohesion (kN/m2)

Friction angle (φ)

20

0

–

0

33

–

–

0

35

15–21

12–23

50

0

100

45

110

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

in the soil during installation. The design of the precast cantilever walls included the analysis of the required penetration depth and the determination of the required pile section from the bending moment in the pile. The Gangweigou floodway was designed with the following conditions/assumptions (Sixth River Management Office, 2011): 1. For global slope stability, two loading conditions, namely, the normal condition and the seismic condition, were considered. 2. The simplified method was adopted to examine the stability of the cantilever wall. 3. Coulomb's earth pressure theory was used to estimate the active and passive earth pressure. It was assumed that the wall friction angle δ = 0.433–0.5ϕ, where ϕ is the friction angle of soil. 4. The required factor of safety for moment was 1.5 for the normal condition and 1.2 for the seismic condition. The factor of safety for penetration depth was 1.2. 5. The surcharges could be idealized as a strip load and resulted in extra horizontal pressure acting on the retaining piles. 6. The groundwater table was 0.5 m higher than the water surface in the floodway because the piles were designed with drainage holes. Because the elevation of the original ground surface varies along the floodway, there were a few different designed sections and pile lengths in the project. From 0K+000 to 0K+120 of the floodway, a single-row cantilever pile wall was designed with the pile length of 9 m. To meet the change in ground elevation, the pile length was adjusted to 12 m from 0K + 120 to 0K + 148. Stability analysis was conducted only for the 12-m pile in the design stage. From 0K + 148 to 0K + 580, the retaining structure was composed of two rows of cantilever pile walls, and the pile length for each row was 12 m. The stability analysis of the retaining structure was conducted separately for the upper-row piles and the lower-row piles. Fig. 2 shows the plane view and typical cross-section of the floodway.

started to deform significantly. By August 12, the flood control road collapsed and sank into the water in the floodway for a length of 20 m. At the outlet of the culvert box (0K+580), the berm in the left bank broke and further expanded by the large flow coming from a joining tributary. Additionally, tensile cracks parallel to the floodway occurred along the slope crest, accompanied by slight tilting of the cantilever walls. On August 13, both the upper row and the lower row of the cantilever walls at the left bank at the outlet of the culvert box tilted increasingly. By August 14, both rows of the retaining structure had completely collapsed. At many places downstream of the culvert, the retaining walls also tilted significantly. The retaining walls near 0K+48 were found tilted, and the backfill soils were flushed out. Tilting and deformation of the downstream retaining walls were observed in several places. Some retaining structures and the cap girder had developed shear fractures, resulting in the backfill loss carried by seepage. During the emergency repair stage, the Sixth River Management Office placed many concrete blocks and riprap on the floodway bank to prevent further damage. Supplementary Material B shows a series of photographs of the above-mentioned failure process. 6. Reexamination of stability Analyses for two failure sections of the case study were conducted to explore the true causes of failure. The wall damage at section 0K+048 took place first; the most severe failure site was at 0K+580. Global stability and wall-penetration depth at these two sites were examined carefully to ascertain the true failure mechanism. Three types of scenarios were considered, namely, 1) the normal condition, 2) the rapid drawdown condition and 3) the condition with riverbed incision, to examine both global stability and wall-penetration depth. According to a field survey after the failure, it was found that the maximum erosion depths of the channel bed were 2.26 m and 2.02 m at 0K + 48 and at 0K+580, respectively.

4.2. Construction method of the cantilever pile walls 6.1. Global failure due to slope sliding The main equipment for the construction of the cantilever walls included cranes, trusses, water jetting devices and pumps. The procedures for the construction of the wall system are briefed as follows. 1. Precast the RC cantilever piles with which a unit with grooves can be connected to its neighboring units. 2. Clear obstacles above and below the ground surface. 3. Install a horizontal positioning girder supported by the completed piles. 4. Use water jets to flush the underlying soil, with the jetting device fixed on a crane truss, to a preset depth. 5. Hang and drive the RC cantilever piles into the preset position. 6. Use the positioning girder, the previously installed pile unit groove and the crane truss to adjust the location of the newly planted pile unit. 7. Flush the bottom soil, and allow the wall unit to be lowered by its own weight to the preset depth. Supplementary Material A shows the photographs of the construction procedures of the wall system. 5. Failure of the Gangweigou floodway Starting on August 7, 2014, a continuous heavy rainfall took place in southern Taiwan, primarily covering the whole Tainan area and causing severe flooding in the Rende and the Annan districts of Tainan. When the floodway was first put to work to bypass the discharge, the estimated peak bypass discharge was 165 m3/s, and the maximum water level reached 8.45 m. On August 10, the damage of the retaining structure was initiated near the concave section of the floodway at 0K + 48. The cantilever wall displaced laterally, and the flood control road on top of the slope

We used the software package GeoStudio (version 2007) to conduct a slope stability analysis involving the global stability of the retaining system. The SLOPE/W module in this package is used for general slope stability analysis based on the two-dimensional limit equilibrium. The SLOPE/W software allows the use of a variety of slice methods to calculate the factor of safety for various sliding surfaces and to locate the sliding surface with the minimum factor of safety. The failure surface can be either a planar or circular surface. Slope geometry, soil strata, groundwater table, pore water pressure distribution, surcharge and reinforcement should all be adequately considered. The input strata strength parameters are listed in Table 1. The simulated conditions include the normal condition and the rapid drawdown condition with various scouring scenarios. For the latter condition, the most critical condition is when the water level in the floodway drops to its lowest level, which was assumed to be 0.5 m above the scour level in various scouring scenarios. The water levels behind the walls were assumed to be at the ground surface. The results of the slope stability analyses are compiled in Fig. 4, Table 2 (for 0K+48) and Table 3 (for 0K+580). For the normal condition and rapid drawdown condition at 0K+48, the factor of safety is high enough to exclude the likelihood of failure. The results of analyses combining rapid drawdown with certain riverbed incision indicated that, even given that the incision depth at 0K+48 had exceeded 3 m, the factor of safety against the global sliding failure would still be greater than 2, and the sliding plane should be located behind the retaining structure, which did not agree with reality. It was hence confirmed that the global stability could not be the failure mechanism at 0K + 48. In section 0K + 580, when considering the rapid drawdown with 2 m of riverbed incision, the factor of safety can be less than 1; however, the failure plane would be behind the flood control roadway, and this again did not agree with what actually

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

111

Fig. 4. The critical sliding planes for global failure: (a) at Section 0K+48; (b) at Section 0K+580.

Table 2 Results of slope stability analyses for global failure (0K+48). Scenarios Conditions

Water level at channel (m)

Water level at backfill (m)

Scour depth(m)

Safety factor

Normal condition Rapid drawdown High water and scour High water and scour High water and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour

4.0 4.0 6.0 6.0 6.0 3.5 3.0 2.5 2.0 1.5

5.6 6.8 6.8 6.8 6.8 6.8 6.8 6.8 6.8 6.8

0 0 1.0 2.0 3.0 1.0 1.5 2.0 2.5 3.0

4.63 4.55 6.11 4.92 4.13 3.62 3.17 2.79 2.51 2.24

112

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

Table 3 Results of slope stability analyses for global failure (0K+580). Scenarios Conditions

Water level at channel (m)

Water level at backfill (m)

Scour depth(m)

Safety factor

Normal condition Rapid drawdown High water and scour High water and scour High water and scour High water and scour High water and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour

4.5 4.5 8.0 8.0 8.0 8.0 8.0 3.5 3.0 2.5

8.0 12.0 12.0 12.0 12.0 12.0 12.0 12.0 12.0 12.0

0 0 1.0 1.5 2.0 2.5 3.0 1.0 1.5 2.0

1.25 1.20 1.36 1.33 1.31 1.28 1.26 1.06 1.01 0.86

happened in the failure case. Hence, global failure was also rejected as the true failure mechanism at 0K+580. 6.2. Stability analysis of the cantilever walls For the stability analyses of the cantilever walls, the distribution of the lateral earth pressure could be properly simplified. The soil profile of the sites is composed of cohesive and cohesionless soils. According to field exploration, the backfill soil under the ground surface at 0K+48 was mainly cohesionless, whereas the soil under the channel bed was mainly cohesive. The loading from the soil behind the upper row of the retaining cantilever wall was simplified as a surcharge. In this work, we idealized the surcharge as a strip load and then converted the surcharge into an extra horizontal force acting on the retaining piles

as illustrated in Fig. 5. For the lower row of the cantilever pile, the effect of the surcharge was ignored. The common formulation of the USA method does not consider the surcharge load and differential water pressures in two sides of the wall. The original USA method was extended in the present work to enable the necessary analysis for this work. As an extension of Figs. 3(c), Fig. 6 illustrates the distribution of net earth pressure and water pressure acting on the wall. The active and passive earth pressures were assumed to follow the Rankine earth pressure theory. Fig. 6(a) and (b) represents cohesionless soil and cohesive soil below the excavation bottom, respectively. The length z in Fig. 6 can be solved from the equilibrium of horizontal forces. In Fig. 6(a), point R is the rotation point of the wall; point O is where the net pressure is zero (i.e., the passive pressure acting towards the right

Fig. 5. Simplification of surcharge effect by converting the surcharge into extra horizontal force acting on the cantilever wall.

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

a)

113

the effective active pressure at the wall base the effective passive pressure at the wall base the resultant force of the active pressures the resultant force of the surcharge loads the resultant force of the water pressure the distance between the excavation bottom and point O the depth of wall penetration measured from the excavation bottom H the excavation depth the depth of the water table hw Y the distance between point O and the wall base Y′ the distance from the resultant force Ra to point O. Y″ the distance from the resultant force RH to the excavation bottom. Z the distance between point G and the wall base c the cohesion of the soil q the surcharge By trial and error, the depth of wall penetration D can be solved from the above equations. The required wall length L is D+H plus a certain safety margin. The safety factor for the penetration depth is defined as the ratio of the actual designed wall length to the required length. The maximum moment is located where the shear force is zero (i.e., the point of zero shear indicated in Fig. 6). The allowable bending moment of the cantilever pile wall is 206 KN-m according to the design specification. The safety factor for the bending failure is defined as the ratio of the allowable bending moment to the maximum moment in the wall. The maximum shear force is located where the net pressure is zero. When the cross section of the cantilever pile is rectangular, the maximum shear stress can be estimated by Eq. (3). The shear strength of concrete can be estimated by an empirical equation such as Eq. (4); in the present case, it is equal to 843 kPa. The factor of safety for shear failure is defined as the ratio of the allowable shear stress to the maximum shear stress.

Pp Pp’ Ra RH Rw a D

b)

τ max ¼ 3V max =2A τC ¼ 0:53 

pffiffiffiffiffi 0 fc

ð3Þ ð4Þ

where

Fig. 6. Illustration of lateral net pressure distribution acting on a cantilever wall: (a) Pressure distribution on granular soil; (b) Pressure distribution on cohesive soil and granular backfill (modified from Cernica (1994)).

is equal to the active pressure acting towards the left). Eqs. (1) and (2) should hold by satisfying the moment equilibrium about the base of the wall. For granular soil:

 0 0 Ra Y þ y þ RH D þ y þ Rw ðD þ ðH−hwÞ=aÞ
   þ PP þ P0P Z2 =4−PP Y2 =6 ¼0

ð1Þ

For cohesive soil:

 0 0 Ra D þ y þ RH D þ y þ Rw ðD þ ðH−hwÞ=aÞ−ð4c−qÞD2 =2
 þ 8∙c Z2 =6 ¼0 where

ð2Þ

τmax the maximum shear stress the allowable shear stress τC the maximum shear force Vmax A the cross-sectional area of the pile fc’ the compressive strength of concrete The factors of safety for penetration depth, maximum moment and maximum shear stress were all reexamined in this work. The results of the analyses are listed in Tables 4, 5 and 6. The results for 0K+48 indicate that, in the case of a high water level in the floodway with a channel bed incision of up to 4 m, the factor of safety for the wall penetration depth would be less than one. Alternatively, for rapid drawdown with a channel bed incision up to 2.5 m, the factor of safety for the wall penetration depth would also be less than one. In those cases, the retaining wall would tilt owing to the imbalance of the active and passive lateral pressures. When the incision depth exceeds 2 m, the factors of safety for moment and shear stress would be less than one; the RC structure would fracture either in bending or in shear. For the section at 0K+580, the factors of safety for the bending moment would be less than one in the upper-row retaining wall during a rapid drawdown, even without any channel bed incision. If the channel bed incision was greater than 1 m, the factors of safety for wall penetration depth, moment and shear stress in the upper-row wall would all be less than one during a rapid drawdown.

114

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

Table 4 Results of the stability analyses of the cantilever wall for various scenarios (0K+48). Scenarios

Conditions Normal condition Rapid drawdown High water and scour High water and scour High water and scour High water and scour High water and scour High water and scour High water and scour High water and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour

Water level at front (m)

Scour Water level at depth backfill (m) (m)

Required penetration depth (m)

Length of piles (m)

Factor of safety for length

4.0

6.0

0

0.83

3.38

2.66

16.10

12.80

169.61

4.53

4.0

6.6

0

1.30

3.85

2.38

40.95

5.03

252.44

3.04

6.6

6.6

0.5

0.73

3.78

2.38

15.44

13.34

87.36

8.79

6.6

6.6

1.0

0.93

4.48

2.01

24.57

8.39

118.36

6.49

6.6

6.6

1.5

1.16

5.21

1.73

36.81

5.60

154.04

4.99

6.6

6.6

2.0

1.41

5.96

1.51

52.71

3.91

194.43

3.21

6.6

6.6

2.5

1.68

6.73

1.34

72.82

2.83

194.43

3.21

6.6

6.6

3.0

1.97

7.52

1.20

97.75

2.11

239.50

2.66

6.6

6.6

3.5

2.31

8.36

1.08

128.15

1.61

289.28

2.23

6.6

6.6

4.0

2.67

9.22

0.98

164.75

1.25

343.75

1.90

3.5

6.6

0.5

1.65

4.70

1.92

60.46

3.41

373.19

2.06

3.0

6.6

1.0

2.15

5.70

1.58

98.41

2.09

503.24

1.53

2.5

6.6

1.5

2.72

6.80

1.32

151.31

1.36

652.71

1.18

2.0

6.6

2.0

3.37

7.95

1.13

220.74

0.93

821.58

0.94

1.5

6.6

2.5

4.10

9.15

0.98

316.88

0.65

1009.86

0.76

As long as the water level in the floodway was as high as the groundwater table behind the retaining wall, all factors of safety for the lowerrow wall would remain greater than one, even if the channel bed incision had been as high as 3 m. However, during a rapid drawdown with a channel bed incision greater than 1 m, the factor of safety for the bending moment and shear stress would be less than one. Once the incision depth was greater than 2.5 m, the factor of safety for the wall penetration depth would become less than one as well.

Maximum bending moment (kN-m)

Factor of safety for moment

Maximum Shear stress (kPa)

Factor of safety for shear

7. Discussion—the causes of failure In this section, the causes of failure at section 0K+048 and section 0K+580 in this case study will be elaborated based on hydrology, hydraulics, failure processes and stability analyses. Comparing the timing of the floodway failures and the water level changes (Fig. 7), it appears that the major failures took place in the low water level periods. The phenomenon is consistent with the results of stability analysis

Table 5 Results of the stability analyses of the upper-row cantilever wall for various scenarios (0K+580). Scenarios

Conditions Normal condition Rapid drawdown Rapid drawdown and scour Rapid drawdown and scour

Water level at front (m)

Scour Water level at depth backfill (m) (m)

Required penetration depth (m)

Length of piles (m)

Factor of safety for length

Maximum bending moment (kN-m)

Factor of safety for moment

Maximum shear stress (kPa)

Factor of safety for shear

4.5

8.0

0

3.83

7.83

1.53

201.49

1.02

439.04

1.75

8.0

12.0

0

5.24

9.24

1.30

272.80

0.76

566.12

1.36

8.0

12.0

0.5

5.87

11.44

1.10

389.16

0.53

707.63

1.09

8.0

12.0

1.0

6.50

12.50

0.96

543.65

0.39

869.25

0.88

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

115

Table 6 Results of the stability analyses of the lower-row cantilever wall for various scenarios (0K+580). Scenarios

Conditions Normal condition High water and scour High water and scour High water and scour High water and scour High water and scour High water and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour Rapid drawdown and scour

Water level at front (m)

Scour Water level at depth backfill (m) (m)

Required penetration depth (m)

4.5

8.0

0

2.34

8.0

8.0

0.5

8.0

8.0

8.0

Factor of safety for length

Maximum bending moment (kN-m)

Factor of safety for moment

6.14

1.95

120.63

1.71

496.13

1.55

1.28

5.58

2.15

44.27

4.65

173.65

4.42

1.0

1.54

6.34

1.89

62.20

3.31

216.38

3.55

8.0

1.5

1.82

7.12

1.69

84.64

2.43

263.81

2.91

8.0

8.0

2.0

2.14

7.94

1.51

112.22

1.81

315.93

2.43

8.0

8.0

2.5

2.48

8.28

1.45

145.63

1.42

372.75

2.06

8.0

8.0

3.0

2.86

9.66

1.24

185.62

1.11

434.26

1.77

4.0

8.0

0.5

2.93

7.23

1.66

184.50

1.12

644.53

1.19

3.5

8.0

1.0

3.62

8.42

1.43

270.09

0.76

812.34

0.95

3.0

8.0

1.5

4.40

9.70

1.24

382.04

0.54

999.56

0.77

2.5

8.0

2.0

5.28

11.08

1.08

525.75

0.39

1206.19

0.64

2.0

8.0

2.5

6.27

12.57

0.95

707.56

0.29

1432.23

0.54

Length of piles (m)

indicating that the safety factor was less than one during the rapid drawdowns of the water level. 7.1. Section 0K+048 The floodway was used to carry the bypass discharge for the first time on August 7, 2014. The floodway at 0K + 048 was located at a

Maximum shear stress (kPa)

Factor of safety for shear

curved section; its left bank is a concave bank where lateral erosion should be more prevalent in principle. The incision of the channel bed could hence take place naturally. As the ground surface at the supporting (passive) side of the cantilever retaining wall was lowered because of incision, the factor of safety of the wall gradually decreased. During the flooding with a high water level in the floodway, the channel bed incision gradually deepened to 2.5 m or more. However, the

Fig. 7. Time histories of rainfall and water level during the flooding event.

116

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

Fig. 8. Sequential processes of failure at section 0K+48.

retaining wall may have still remained stable as long as the high water level in the floodway channel was able to balance the pore water pressure and lateral stress behind the wall. When the peak discharge was over, the water level in the floodway would quickly fall, shortly resulting in a rapid drawdown situation in the bank. Under the rapid drawdown condition with the channel bed incision exceeding 2.5 m, the shear stress and the bending moment in the wall would exceed the allowable strength and cause fractures of the wall. With a high water pressure in the backfill, silty sand was carried out by the seepage with a high hydraulic gradient and formed holes and caves (as Fig. 8a shows). Initially, these holes and caves were not notable; gradually, they expanded and enlarged, finally forming into a drainage route and causing the total collapse of the flood control roadway on top of the slope on August 10 (Fig. 8b). On August 11–12, another heavy rainfall in the upstream catchment caused a new cycle of the rising and falling of the water level in the floodway. This rapid drawdown cycle worsened the situation: the cap girder was completely broken owing to the excessive shear force within the wall structure. After the channel bed incision, the wall penetration

depth was no longer sufficient, so the wall tilted toward the floodway. The soil behind the wall was exposed and was eroded as a result (Fig. 8c and d). 7.2. Section 0K+580 Section 0K+580 was located at the junction of the culvert boxes and the open channel. Owing to the difference in roughness, velocity change and erosion were likely to take place there. The channel bed was not protected by any ground sill. During the first bypass flow on August 7, there had been some local erosion in this section. Although the retaining wall at that time appeared stable, the wall actually had tilted slightly, and there were already some observable tensile cracks in the slope crest (Fig. 9a). On August 12, the flow discharge became very large, such that the channel bed erosion accelerated. The cantilever retaining wall tilted more severely; the tensile crack on the slope had fully expanded. In the meantime, a drainage channel connecting to the floodway was blocked by various obstacles. As a result, the surface runoff flowed

Fig. 9. Sequential processes of failure at section 0K+580.

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

over the bench of the slope and raised the water pressure on the active side. As a response, the lower-row cantilever retaining wall tilted and broke. The runoff flow damaged the slope bench and eroded the soil away (Fig. 9b). By August 13, soil loss deteriorated increasingly, and the upper-row retaining wall lost it supporting capability and collapsed completely. Afterwards, the active lateral stress, together with the loading from the displaced slope, had to be transferred to the lower-row retaining wall (Fig. 9c). On August 14, the water level in the floodway dropped after the heavy rainfall stopped. The retaining walls on both sides of the floodway failed during the rapid drawdown (Fig. 9d). 8. Lessons from the failure case Likely because of the relatively mild slope of the floodway in this case, the potential for the incision of the river bed was not considered in the design without a quantitative assessment in advance. Consequently, neither ground sill nor energy dissipation measures was installed to protect the floodway bed from erosion. Nevertheless, the observed maximum erosion depths of the channel bed were 2.26 m and 2.02 m at 0K+ 48 and at 0K+ 580, respectively, according to the field investigation after failure. Under a high flow velocity during the peak discharge, the river bed may not be able to resist the erosion from the shear stress on the bed. Fig. 7 compiles the hourly rainfall and estimated water level during the long-lasting rainfall event. According to the recorded data, the accumulative rainfall was 535 mm, approximately corresponding to a return period of 10 years. A hydraulic analysis using the software program HEC-RAS was conducted to estimate the hydraulic condition and possible incision depth under a maximum bypass discharge of 165 m3/s. The results of the hydraulic analysis indicated that the flow velocity could reach 3.35 m/s at 0K+580 with a water level of 8.45 m and an energy gradient of 0.245%. At 0K+48, the flow velocity could reach 2.18 m/s with a water level of 7.34 m and a very small energy gradient of 0.058%. With the method proposed by Briaud et al. (2011), one can estimate the ultimate incision depth from these hydraulic variables. The estimated ultimate incision depth at 0K + 48 and 0K + 580 under the corresponding flow conditions are 3.12 m and 3.31 m, respectively. Some simple type of ground sill work or revetment could help protect the riverbed of the floodway from erosion and avoid the incision that occurred in the case. Moreover, the stability of the retaining system was not examined for the rapid drawdown condition in designing the walls. Unfortunately, the rapid drawdown condition with the floodway incision was found to be the most critical condition for the stability of the retaining system, according to the analysis in the previous sections. The riverbed incision was not expected or considered in the design. As a result, the designed penetration depth became insufficient after the channel incision in the floodway during rapid drawdown. As mentioned, the combined condition of rapid drawdown with a significant channel bed incision can explain why the retaining system would fail in the low-water-level stage during the hydrological event. The flow velocity near 3 m/s was not considered to be very high in a typical river channel in Taiwan. Briaud et al. (2008) proposed a classification chart for characterizing the erodibility of geomaterials; this chart also correlates the range of erosion rate and flow velocities for various soils and rock masses. According to the classification chart, the erodibility class of the in situ soil is from medium (for CL) to very high (for SM); in addition, the erosion rate of the soil formation could be very high for flow velocity higher than 3 m/s. With the channel incision, the water-head difference in the active side and the passive side of the retaining wall further increased during the rapid drawdown stage while the water level in the floodway significantly dropped. Accordingly, the unbalanced water pressure and the reduction of passive pressure because of bank erosion further enhanced the movement of the retaining walls toward the

117

floodway. The retaining system failed as a result. The failure of the case could have been avoided if the scenario of the channel bed incision and rapid drawdown had both considered in the design. The precast walls were installed by the jetting-assisted method. Compared with impact driving with a hammer, the water jet method is an effective means for driving concrete sheet piles into the strata of dense sand or hard clay or soil containing some cobbles (Tsinker, 1988). However, the in situ soil structure is likely significantly altered owing to water jetting; the large hydraulic gradient during jetting and the induced seepage flow will very likely affect the original soil characteristics (Shepley and Bolton, 2014). The shear strength parameters in Table 1 were the laboratory test results of the specimens without the influence of jetting. The actual shear strength of the in situ soil at failure is unclear. However, the frictional resistance of the soil adjacent to a jetting-driven pile may be lower than that of an impact-driven pile (Tsinker, 1988). It may be advisable to be more conservative in design when precast walls would be installed by the jetting-assisted method. The designer may consider setting a slightly higher required factor of safety than the usual case for conservative concern. 9. Concluding remarks The Gangweigou floodway was designed to bypass partial discharge in the Gangweigou River during heavy rainfall events. This floodway was an excavated open channel supported by single or double rows of precast cantilever pile walls on each side of its banks. The original design did not consider the possible channel bed incision at all. The stability of the retaining structure was examined based only on the channel bed elevation. Owing to the mild slope of the floodway, no ground sill was considered necessary. However, for the failure sections at 0K+048, the lateral erosion was notable on the concave bank. The significant channel bed incision resulted in the shortage of enough penetration depth and the insufficiency of the supporting capacity of the cantilever pile wall on the passive side, causing the structural failure of the wall. At the other site of the severe failure, 0K+580, the intense erosion was due to the inconsistent roughness in the junction of the culvert box and the open channel. The velocity variation near the junction resulted in excessive erosion and indirectly triggered the failure of the retaining system of the channel. In both sections of the failure sites, it appears that ignoring the possible bed incision or bank erosion was likely the primary cause of the improper design. Furthermore, a high water level in the floodway was not the critical condition for controlling the stability of the flood control wall. Instead, the most critical condition was when the water level in the floodway quickly dropped, which resulted in a rapid drawdown in the channel banks. All precast cantilever pile walls in this project were driven by the jetting-assisted method. It is expected that the surrounding soil subjected to flushing would have been seriously disturbed, despite the difficulty of quantitative evaluation. The field strength of the soil after flushing was likely weaker than the tested sample obtained through field exploration sampling. All analyses in the foregoing context have assumed the strength parameters obtained from laboratory tests. Thus, the actual factor of safety in the field in reality could be even smaller than what was calculated and reported in this paper. Nevertheless, the analyses of the case study may facilitate understanding of the reasons and processes of the failure case. Acknowledgments The authors retrieved a variety of design drawings, specification, photographs and material parameters related to the Gangweigou floodway from various reports of the Water Resources Agency (WRA), Taiwan. The sixth River Management Office, WRA is acknowledged. The authors also wish to acknowledge the following contributors: 1) Jui-

118

M.-H. Cheng et al. / Engineering Geology 209 (2016) 106–118

Hsiang Wu, the section chief in the First River Management Office, who provided photographs during construction of the cantilever walls; 2) Shin-Hsiung Lin, an engineer in the Water Resources Agency, who provided photographs during the completion and failure of the floodway; and 3) His-Ben Shu, an engineer in the Water Resources Planning Institute, who provided the results of the HEC-RAS simulation. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.enggeo.2016.05.014. References Alejo, G.T., 2013. Numerical Analysis of Cantilever and Anchored Sheet Pile Walls at Failure and Comparison with Classical Methods (Master thesis) Department of Civil Engineering, Universitat Politècnica de Catalunya, BarcelonaTech, pp. 10–13. Bica, A.V.D., Clayton, C.R.I., 1989. Limit Equilibrium Design Methods for free Embedded Cantilever Walls in Granular Materials. ICE Proceedings, Part 1 Vol. 86, pp. 879–898. http://dx.doi.org/10.1680/iicep.1989.3161. Bolton, M.D., Powrie, W., 1987. The collapse of diaphragm walls in clay. Geotechnique 37 (3), 335–353. http://dx.doi.org/10.1680/geot.1987.37.3.335. Bourdouxhe-Barnich, M.-P., Piault, D., Ursat, P., Herve, S., 2002. Jetting-Assisted Sheet Pile Driving. Ninth International Conference on Piling and Deep Foundations, Nice, France, Article No. 1020. Bowles, J.E., 1988. Foundation Analysis and Design. fourth ed. McGraw-Hill Book Co., New York. Briaud, J.L., Chen, H.C., Govindasamy, A.V., Storesund, R., 2008. Levee erosion by overtopping in New Orleans during the Katrina hurricane. J. Geotech. Geoenviron. 134, 618–632. http://dx.doi.org/10.1061/(asce)1090-0241(2008)134:5(618). Briaud, J.L., Chen, H.C., Chang, K.A., Oh, S.J., Chen, S., Wang, J., Li, Y., Kwak, K., Nartjaho, P., Gudaralli, R., Wei, W., Pergu, S., Cao, Y.W., Ting, F., 2011. The Sricos — EFA Method Summary Report. Texas A&M University.

Cernica, J.N., 1994. Geotechnical Engineering: Foundation Design. John Wiley and Sons, New York. Chen, Y.G., 1992. Sea-Level Change and Neotectonics in Southern Part of Taiwan Region since Late Pleistocene (Ph.D. Dissertation) Department of Geosciences, National Taiwan University, Taiwan (in Chinese). CMIWAC, 2014. Seawall Owner's Manual. Waterways Advisory Committee, City of Marco Island, Florida, US. Day, R.A., 1999. Net pressure analysis of cantilever sheet pile walls. Geotechnique 49 (2), 231–245. http://dx.doi.org/10.1680/geot.1999.49.2.231. Dutta, P.K., Vaidya, U., 2003. A Study of the Long-Term Applications of Vinyl Sheet Piles. US Army Corps of Engineers, Engineer Research and Development Center, Cold Regions Research and Engineering Laboratory. GeoStudio (Version 2007) [Software], 2007. GEO-SLOPE International LTD. http://www. geo-slope.com/. Hameed, R.A., Gunaratne, M., Putcha, S., Kuo, C., Johnson, S., 2000. Lateral load behavior of jetted piles. Geotech. Test. J. 23 (3), 358–368. King, G.J.W., 1995. Analysis of cantilever sheet-pile walls in cohesionless soil. J. Geotech. Eng. ASCE 121, 629–635. Liu, J., 2002. Centrifugal Modeling of Multi-Braced and Unbraced Excavation Failures. Physical Modelling. 8th International Conference on Physical Modelling in Geotechnics, St Johns, Canada, pp. 841–846. Padfield, C.J., Mair, R.J., 1984. Design of Retaining Walls Embedded in Stiff Clay. CIRIA Report 104. Shepley, P., Bolton, M.D., 2014. Using water injection to remove pile base resistance during installation. Can. Geotech. J. 51 (11), 1273–1283. http://dx.doi.org/10.1139/cgj2013-0240. Sixth River Management Office, 2011. The Detail Design of the Gangweigou Floodway (in Chinese). Tsinker, G.P., 1988. Pile jetting. J. Geotech. Eng. 114 (3), 326–334. US Army Corps of Engineers, 1994. Engineering and Design of Sheet Pile Walls, EM 11102-2504, Washington, DC. Water Resources Agency, 2010. The Exploration Report for the Design and Construction of the Gangweigou Floodway (in Chinese).