9
Forensic numerical analysis of gas venting in Southwestern Ontario R. Kerry Rowe a, , Ahmed Mabrouk b, 1 a Geotechnical and Geoenvironmental Engineering, GeoEngineering Centre at Queen's-RMC, Department of Civil Engineering, Queen's University, Kingston, ON, Canada K7L 3N6 b GeoEngineering Centre at Queen's-RMC, Department of Civil Engineering, Queen's University, Canada abstract article info Article history: Received 24 October 2011 Received in revised form 13 August 2012 Accepted 6 September 2012 Available online 17 September 2012 Keywords: Hydrofracturing Excavation Finite element Gas venting Gassy soil A forensic modeling study examines the potential causes of gas and water venting observed during the excava- tion of a landll in a 40 m thick clayey till deposit in southwestern Ontario, Canada. The clayey till is known to be underlain by permeable, natural gas bearing, rock and gas has been diffusing through the clayey deposit over about the last 13,00015,000 years. Attention is focused on 2D modeling of hydrofracturing as well as gassy soil behavior. The model examines the predicted location of venting with and without the presence of gassy sand lenses and a discontinuous layer of basal till between the bedrock and the low permeability till. The model- ing shows that hydrofracturing alone can explain the occurrence and location of the observed venting in one of the sub-cells. In the other sub-cell, hydrofracturing explains the occurrence but not the location of the observed venting or the presence of the sand in the venting water at this location. Modeling of appropriately placed gassy sand lenses is shown to explain the occurrence and location of the observed venting as well as the presence of both waters with different geochemistry from the bedrock and the till in the venting waters. © 2012 Elsevier B.V. All rights reserved. 1. Introduction A landll was constructed in low permeability clayey till underlain by a bedrock aquifer (Figure 1) using a progressive excavation and ll- ing technique. The landll was divided into cells and smaller sub-cells. The construction procedure involved gradually excavating selected areas of a cell forming an overall initial plateau at about elevation 186 mASL (i.e., 14 m below the original ground surface elevation of 200 mASL; Figure 2) followed by excavation of a secondary plateau to 182 mASL. The secondary plateau was a platform from which 6 m deep trenches (35 m×35 m) were excavated (to 176 mASL) and then lled with waste not long after being excavated. As the excavation in adjacent trenches (denoted as Sub-Cells 1 and 2 herein) approached its maximum design depth of 24 m, venting of gas (mainly methane) through the clayey till was observed at three locations denoted as Seeps A, B and C. Work was halted around the affected zone and precau- tions were taken to keep that area isolated from waste. No groundwater contamination resulted from this event. A previous paper (Mabrouk and Rowe, 2011) modeled the effect of sand lenses where gas exsolu- tion occurred (a) prior to the excavation, or (b) during the excavation. Both approaches lead to a similar conclusion that while liquefaction of a sand lens near the base of the excavation may have contributed to the formation of the seeps and vents, the exsolution of gas in these sand lenses during unloading could not explain the level of venting that was observed in the eld. Thus the objective of this present paper is to examine other potential explanations for the venting of gas and water at this site. 2. Geological conditions and stratigraphy A silty clay layer, known as St Joseph Till, extends for 14 m (200186 mASL) from soil surface (Figure 1). This unit is weathered, fractured and hydraulically active for the top 6 m below which it is an intact, clay rich, low hydraulic conductivity (typically less than ~8×10 -10 m/s) aquitard for the bottom 8 m. The St Joseph Till is underlined by a 2226 m (down to 164160 mASL) of clayey silt shale till of low hydraulic conductivity (typically less than ~2 × 10 -10 m/s) known as the Black Till. The Black Till is regarded as an unusual deposit as it is normally consolidated at depth (in contrast to most glacial origi- nated tills) except for the top 2 m, denoted as Black Till 1, which is slightly over-consolidated due to presumed water table uctuations post deposition (Quigley and Ogunbadejo, 1976). Sand and silty sand pockets of 1.1 m average thickness are encountered at different eleva- tions within the Black Till and particularly in a zone about 2428 m below ground surface (172176 mASL). The Black Till and the bedrock are sometimes separated by a thin, weak, discontinuous layer of sandy clayey silt, known as the Basal Till. However, it appears that this layer is missing below much of the affected zone (Rowe and Mabrouk, 2007). The shale bedrock, known as Kettle Point Formation, has a rela- tively high hydraulic conductivity (typically ~ 1 × 10 -5 m/s) and is an aquifer. The Kettle Point Formation is known elsewhere as the Antrium Shale (in Michigan) and the Ohio Shale (in Ohio) and is believed to be both a source and reservoir of methane(Martini et al., 1998). The Engineering Geology 151 (2012) 4755 Corresponding author. Tel.: +1 613 533 3113; fax: +1 613 533 2128. E-mail addresses: [email protected] (R.K. Rowe), [email protected] (A. Mabrouk). 1 Tel.: +1 613 533 3113; fax: +1 613 533 2128. 0013-7952/$ see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.enggeo.2012.09.005 Contents lists available at SciVerse ScienceDirect Engineering Geology journal homepage: www.elsevier.com/locate/enggeo

Forensic numerical analysis of gas venting in Southwestern Ontario

Embed Size (px)

Citation preview

Engineering Geology 151 (2012) 47–55

Contents lists available at SciVerse ScienceDirect

Engineering Geology

j ourna l homepage: www.e lsev ie r .com/ locate /enggeo

Forensic numerical analysis of gas venting in Southwestern Ontario

R. Kerry Rowe a,⁎, Ahmed Mabrouk b,1

a Geotechnical and Geoenvironmental Engineering, GeoEngineering Centre at Queen's-RMC, Department of Civil Engineering, Queen's University, Kingston, ON, Canada K7L 3N6b GeoEngineering Centre at Queen's-RMC, Department of Civil Engineering, Queen's University, Canada

⁎ Corresponding author. Tel.: +1 613 533 3113; fax:E-mail addresses: [email protected] (R.K. Rowe

(A. Mabrouk).1 Tel.: +1 613 533 3113; fax: +1 613 533 2128.

0013-7952/$ – see front matter © 2012 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.enggeo.2012.09.005

a b s t r a c t

a r t i c l e i n f o

Article history:Received 24 October 2011Received in revised form 13 August 2012Accepted 6 September 2012Available online 17 September 2012

Keywords:HydrofracturingExcavationFinite elementGas ventingGassy soil

A forensic modeling study examines the potential causes of gas and water venting observed during the excava-tion of a landfill in a 40 m thick clayey till deposit in southwestern Ontario, Canada. The clayey till is known to beunderlain by permeable, natural gas bearing, rock and gas has been diffusing through the clayey deposit overabout the last 13,000–15,000 years. Attention is focused on 2D modeling of hydrofracturing as well as gassysoil behavior. The model examines the predicted location of venting with and without the presence of gassysand lenses and a discontinuous layer of basal till between the bedrock and the low permeability till. Themodel-ing shows that hydrofracturing alone can explain the occurrence and location of the observed venting in one ofthe sub-cells. In the other sub-cell, hydrofracturing explains the occurrence but not the location of the observedventing or the presence of the sand in the venting water at this location. Modeling of appropriately placed gassysand lenses is shown to explain the occurrence and location of the observed venting as well as the presence ofboth waters with different geochemistry from the bedrock and the till in the venting waters.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

A landfill was constructed in low permeability clayey till underlainby a bedrock aquifer (Figure 1) using a progressive excavation and fill-ing technique. The landfill was divided into cells and smaller sub-cells.The construction procedure involved gradually excavating selectedareas of a cell forming an overall initial plateau at about elevation186 mASL (i.e., 14 m below the original ground surface elevation of200 mASL; Figure 2) followed by excavation of a secondary plateau to182 mASL. The secondary plateau was a platform from which 6 mdeep trenches (35 m×35 m) were excavated (to 176 mASL) and thenfilled with waste not long after being excavated. As the excavation inadjacent trenches (denoted as Sub-Cells 1 and 2 herein) approachedits maximum design depth of 24 m, venting of gas (mainly methane)through the clayey till was observed at three locations denoted asSeeps A, B and C.Workwas halted around the affected zone and precau-tionswere taken to keep that area isolated fromwaste. No groundwatercontamination resulted from this event. A previous paper (Mabroukand Rowe, 2011) modeled the effect of sand lenses where gas exsolu-tion occurred (a) prior to the excavation, or (b) during the excavation.Both approaches lead to a similar conclusion that while liquefaction ofa sand lens near the base of the excavation may have contributed tothe formation of the seeps and vents, the exsolution of gas in thesesand lenses during unloading could not explain the level of venting

+1 613 533 2128.), [email protected]

rights reserved.

that was observed in the field. Thus the objective of this present paperis to examine other potential explanations for the venting of gas andwater at this site.

2. Geological conditions and stratigraphy

A silty clay layer, known as St Joseph Till, extends for 14 m(200–186 mASL) from soil surface (Figure 1). This unit is weathered,fractured and hydraulically active for the top 6 m belowwhich it is anintact, clay rich, low hydraulic conductivity (typically less than~8×10−10 m/s) aquitard for the bottom 8 m. The St Joseph Till isunderlined by a 22–26 m (down to 164–160 mASL) of clayey silt shaletill of low hydraulic conductivity (typically less than ~2×10−10 m/s)known as the Black Till. The Black Till is regarded as an unusual depositas it is normally consolidated at depth (in contrast to most glacial origi-nated tills) except for the top 2 m, denoted as Black Till 1, which isslightly over-consolidated due to presumed water table fluctuationspost deposition (Quigley and Ogunbadejo, 1976). Sand and silty sandpockets of 1.1 m average thickness are encountered at different eleva-tions within the Black Till and particularly in a zone about 24–28 mbelow ground surface (172–176 mASL). The Black Till and the bedrockare sometimes separated by a thin, weak, discontinuous layer of sandyclayey silt, known as the Basal Till. However, it appears that this layeris missing below much of the affected zone (Rowe and Mabrouk,2007). The shale bedrock, known as Kettle Point Formation, has a rela-tively high hydraulic conductivity (typically ~1×10−5 m/s) and is anaquifer. The Kettle Point Formation is known elsewhere as the AntriumShale (in Michigan) and the Ohio Shale (in Ohio) and is believed to be“both a source and reservoir of methane” (Martini et al., 1998). The

Fig. 1. Soil profile.

48 R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

whole area of southwest Ontario has been subjected to oil and gas explo-rations and production since 1800's with the first recorded oil well inNorth America (Oil springs: Dittrich et al., 2010) located about 30 kmfrom site. The bedrock is encountered at a depth of between 37 and42 m below the ground surface (163–158 mASL) with the peak eleva-tion (162.4–163 mASL) being located near/below, where the ventingwas observed. Hence, this zonewould be a suitable location for gas accu-mulation at the interface of permeable bedrock and low permeabilityBlack Till. The water table is very close to ground surface and initialpore pressure distribution was hydrostatic.

3. Problem description

The following description is based on unpublished reports and thesenior author's personal observations at the site. During excavation ofSub-Cell 2 and concurrent filling of Sub-Cell 1 with waste (after it hadbeen excavated to its lowest elevation of 176 mASL), gas and waterventing was observed emanating from a narrow elongated pocket ofsand containing minor gravel (Seep A) between the two sub-cells(Figure 2). The erosive action of gas andwater formed a tube like struc-ture about 0.1 m in diameter for a limited distance (b1 m). Initial waterflow was about 8 l/min (0.13 l/s) with continuous gas bubbling. Theflow declined to less than 1 l/min (0.017 l/s) within 10 days.

Twoweeks later as the excavation proceeded in Sub-Cell 2, multiplevents (called Seep B) were observed at base of the sub-cell (elevation

Fig. 2. Soil profile and cross section of the excavation without fill (the end of excavation). Cr

176 m ASL). Seep B was comprised of two prominent vent openingsSeep B-NE and Seep B-SW each with flow between 5 and 10 l/min(0.08–0.17 l/s) as well as some secondary openings which had flowsof less than 1 l/min (0.017 l/s). Gas flow could not be estimated butappeared to be continuous at all vent openings. The vents wereinterconnected. At Seep B-NE the gas and water were dischargingfrom a rounded (caused by the erosive action of the gas/water) featuredipping to the south at about 45°. Dendritic cracks were observed par-allel and close to the primary opening. Seep B-SW was located 5 mfrom the berm toe. Water and gas venting, while significantly reduced,was still evident 4 months later.

Four days after flow initiation from Seep B, venting of gas and waterwas observed at the edge of Sub-Cell 1(denoted as Seep C). Water flowwas at a rate of less than 0.5 l/min (0.008 l/s) from the intersection ofsmall scale cracks.

Minor bubbling was observed at other locations within Sub-Cells 1and 2 where water had accumulated following a rain event. On a clos-er inspection, bubbles appeared to be emanating from small openings(small tension cracks in the exposed clay).

Flow was stopped from all seep vent openings (A, B and C) by theconstruction of a 5 m thick compacted clay liner over the area.

4. Site hydrogeology

Ground water collected from wells in the active aquitard wasenriched in sulphate, calcium, andmagnesium, andwas depleted in po-tassium, sodium and chloride. However, ground water samples fromthe interface aquifer were enriched in chloride and sodium, and weredepleted in calcium and sulphate. Water samples from Seeps A, B andC had high sodium and chloride concentrations suggesting water fromthe lower aquifer. The sulphate concentration in some samples fromSeeps A and C was higher than the typical values for aquifer water,suggesting that the water source was a mix of aquitard and aquiferpore fluid.

5. Numerical model and parameters

The finite element software, Abaqus, was used to examine the con-ditions that prevailed during excavation. An elasto-plastic model witha Mohr Coulomb failure criteria and a non‐associated flow rule withzero dilatency angle was adopted to conduct an effective stress analysisfor the site. The elastic behavior was taken to be linear and isotropic. An8-noded element with biquadratic displacement function, bilinear porepressure and reduced integration was used for the 2D modeling. Theelement has three degrees of freedom (ux, uy, and pore pressure at thecorners). The average element size was 2 m square.

The initial stress conditions in the key elevations in the Black Tillcorresponded to Ko conditions for a normally consolidated soil. Thesestresses were established in the model by subjecting to soil to gravity

oss section passes through Seeps A and B. Seep C is out of the plane of the cross-section.

Table 1Unit weight and shear strength parameters (based on unpublished reports).

Depth Layer γ (kN/m3) c' (kPa) Φ' (°)

200–194 St Joseph Till 1 21.5 24 25194–186 St Joseph Till 2 21 16 27186–184 Black Till 1 20.5 24 24184–162 Black Till 2 19.6 24 24162–164 Basal Till (where present) 18.2 9 18b162 Bedrock 23 40 40

Table 2Additional soil properties.

Layer E'(kPa) ν(−) e0(−) k (m/s)

St Joseph Till 1 55,000 0.4 0.47 8.3×10−8

St Joseph Till 2 55,000 0.4 0.49 8.3×10−10

Black Till 1 55,000 0.4 0.56 2.4×10−10

Black Till 2 30,000 0.4 0.74 2.4×10−10

Basal Till (where present) 15,000 0.4 1.2 1.0×10−8

Bedrock 200,000 0.4 0.25 1.0×10−5

49R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

forces under conditions of zero lateral strain. The subsequent excava-tion was modeled through a series of steps by deactivation andreactivation of elements to simulate the construction sequence on site.Accordingly, the model geometry changed with time to follow the var-iations in site contours. The stresses at the end of any one stepwere theinitial stresses for the next step in the analysis.

The 2D modeling reported herein examined a cross section passingthrough Seeps A and B. Fig. 2 represents the final excavated sub-cell ge-ometry (without the fill). Seep C shown in Fig. 2 is out of the plane of thesection and is not modeled in this paper.

The base boundary was taken to be rough and rigid (ux=uy=0,where ux and uy are the displacement in x and y directions respectively),while the lateral boundarieswere taken to be smooth and rigid (ux=0).The initial ground surface and the subsequent surfaces exposed by theexcavation were assigned zero pore water pressure throughout the ex-cavation. Hydrostatic pore water pressures were specified at the farfield lateral boundaries. A zero flux boundary was specified at themodel base at a depth of 6 mbelow the upper fractured bedrock aquifer(Figure 2).

The effective shear strength parameterswere selected based on fieldand laboratory data. The properties adopted in the analyses are summa-rized in Table 1. The soil stiffness and hydraulic conductivity parameters(Table 2) were selected based on Dittrich (2000) who studied slopemovement during excavation at a nearby railway cutting in the samegeological units.

An initial analysis was performed using the standard commercial FEcodewithout consideration of hydrofracturing. Unloadingwas simulateduntil the design excavation elevation (176 mASL) of Sub-Cell 2 was

Fig. 3. 2D finite element mesh used of 2D plane strain analysis showing tensile zone after futhe boxes denoted as “Section A” and “Section B”.

reached. At this time tension (in terms of effective stress) had developedat some locations at the interface between the Black Till and the bedrockaquifer (Figure 3). Although these local tensile zones didn't affect theoverall stability of the excavation and were not extensive enough to ex-plain the formation of seeps that were actually observed, they suggestthe potential for initiation of hydrofracturing at the interface betweenthe bedrock and till. As a consequence, the FE code was modified to ap-proximately account for hydrofracturing as described below.

6. Hydro-fracturing modeling

Fractures can be classified based on their cause, into tensile and shearmode cracks (Van der Pluijm and Marshak, 2004); this paper deals withthe former. Hydrofracturing is frequently used inwater and oil wells as atechnique to increase well productivity within layers of low hydraulicconductivity (e.g., clay). Furthermore, the hydrofracturing technique isused for enhancing the environmental remediation performance of pol-luted sites with fine grained soil layers (Murdoch, 1988, 1989, 1995,2000; Murdoch and Slack, 2002). In these cases, pore pressures are in-creased until the effective stress in the minor principal stress directionexceeds the soils' strength. This will lead to the development of tensioncracks which will propagate and in the case of oil recovery, increasethe hydraulic conductivity to a point where there is improved oil recov-ery. If this occurs near the soil surface the tension crack may continueuntil it reaches that surface if the injected volume is sufficientlyhigh. Hydrofracturing is divided into two stages: crack initiation andpropagation. The pressure required to initiate a hydrofracture dependson many factors including the initial stresses (circumferential stress incase of a circular well geometry), Poison's ratio (Bjerrum et al., 1972)and water content (Murdoch, 1993a). The pressure required to initiatehydrofracturing increases with increasing total minor principal stress(Lo andKaniaru, 1990) and hence for clayey soil, the degree of consolida-tion has a significant effect on the onset of hydrofracturing. Crack initia-tion is independent of soil stiffness (Bjerrum et al., 1972) howeverstiffness affects crack propagation (Brenner and Gudmundsson, 2004).In soil, the hydrofracture propagation pressure has not been investigatedto the same extent as the hydrofracture initiation pressure. This islikely because most studies have been concerned with preventinghydrofracturing which depends on limiting the pressure to below theinitiation pressure (Murdoch, 1993a). The hydrofracture propagationpressure generally is less than the initiation pressure (Lo and Kaniaru,1990).

Murdoch (1993a, 1993b, 1993c) presents an explanation of thehydrofracturing formation mechanism in soil that can be summarizedas follows. The fracture is nucleated by high pore pressure when favor-able conditions prevail at a specific location. Porefluid either propagatesthrough the fracture tip or infiltrates through the surrounding soilforming different hydrofracture surfaces (starter slot, parent fracture,lobes, leading edge). Distribution of fluid pressure through the fracturecontrols the propagation mechanism (Murdoch, 1993a).

ll excavation of Sub-Cell 2. Several future figures will show results for the region inside

50 R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

The crack propagation pressure and direction of propagation dependon many variables including soil anisotropy, heterogeneity, etc.(Brenner and Gudmundsson, 2004). However, for partially to fully sat-urated silty clay—which is the case studied herein, the propagation offractures is normal to the direction of the minimum principle compres-sive stress (Murdoch, 1993a; Chang, 2004).

Most of the theoretical formulae predicting the crack initiation pres-sure depend on empirical values (Bjerrum et al., 1972; Jaworski et al.,1981) or laboratory parameters (Murdoch, 1993a, 1993b, 1993c;Murdoch and Slack, 2002; Chang, 2004) or both (Panah andYanagisawa, 1989; Lo and Kaniaru, 1990). It was not practical to usethese formulae in analyzing this problem. The theory of expansion ofa cylindrical cavity states that for tensile failure, a crack is assumed tobe initiated if the effective circumferential stress (σ'θ) exceeds soil ten-sile strength (σ't) (Eq. (1)). Soil tensile strength is usually very low andis neglected in many studies (Bjerrum et al., 1972; Massarsch, 1978;Panah and Yanagisawa, 1989). Likewise in this study the tensilestrength is assumed to be zero and hence it is assumed that cracks(hydrofractures) are initiated when the minimum effective stressdrops to zero (i.e., the fluid pressure is equal to the minor principaltotal stress and hence there is imminent tension).

σ ′θ þ Δσ ′

θ > σ t : ð1Þ

Modeling crack propagation has been the subject of many papers(see Ingraffea, 2008 for a review). Some approaches use a geometricalrepresentation to simulate crack propagation. In this approach, cracksare explicitly represented in the geometrical model such as constrainedshapemethods (e.g., prescribedmethods (Ortiz and Pandolfi, 1999) andanalytical geometry methods (Wang et al., 1997)) and arbitrary shapemethods (e.g., mesh free method (Belytschko et al., 1996) and adaptiveFEM/BEM methods (Martha et al., 1992)). Another common approachto model crack propagation is by implicitly acknowledging the cracks inthe numerical model (e.g. constitutive methods: smeared crack (Weiheet al., 1998), element extinction (Beissel et al., 1998) and kinematic/enriched element methods (Belytschko and Black, 1999)).

Linear elastic fracture mechanics (LEFM) commonly used to ap-proximate crack propagation in brittle and quasi-brittle materialshave been successfully used within the past few decades to modelfracture propagation in geotechnical problems (Sture et al., 1999;Wang et al., 2007). LEFM uses critical stress intensity (KI) which is afunction of loading, fracture size and structural geometry, as a criteri-on for fracture propagation. A material property “toughness” is intro-duced to represent the material resistance to a mode I fracture (themost common fracture type with the crack plane normal to the direc-tion of largest tensile loading). If the hydrofracture length is known,soil toughness can be used as an indicator of the driving pressure re-quired for the hydrofracture propagation (Murdoch, 1993b). For clay,soil toughness is influenced by many variables (water content, degreeof consolidation, dry density and other factors—Murdoch, 1993b;Wang et al., 2007), and its value is obtained by soil testing (manytesting methods have been proposed to measure soil toughness—Zhu and Joyce, 2012).

Commercial codes that do coupled fluid flow propagation of curv-ing cracks, as considered here, are not available and existing researchcodes are not readily suitable to the analysis of the problem consid-ered and, even if they were, they would be of little use to practitionersexploring practical problems such as that examined herein. Thus a de-cision was made to modify existing commercial software (Abaqus) toboth facilitate development and provide an approach that could beused by others familiar with the widely used program.

This study adopted an effective media approach to simulatehydrofracture growth. The FE code was modified to approximatelyaccount for hydrofracture propagation by significantly increasing hy-draulic conductivity (by three orders of magnitude) in regions wherethe effective stress dropped to zero. The adopted approach treated the

problem in the spirit of an effective mediumwithout treating the crackexplicitly. This approach was inspired by element failure algorithm(element extinction) for modeling crack propagation where elementsare deleted from the mesh when the cracking criterion is reached(Beissel et al., 1998). The adopted effective medium approach allowedthe simulation of important aspects of the problem—propagationcriteria, increment of crack growth, recalculation of stresses, propertychange during crack extension and coupling to fluid flow, with reason-able development time (as needed to study different scenarios of theconditions surrounding the excavation).

The approach adopted for the analysis presented in this paper cre-ates a path for pore pressures to develop along planes of weaknesswhere hydrofracturing is considered to occur. The approach doesnot require remeshing or modifying the element surface duringhydrofracture propagation. The model is conservative in that it as-sumes that the propagation pressure is equal to the initiationpressure.

7. Sub-Cell 1

A 2D plane strain analysis was conducted for the section shown inFig. 2 using the version of the FE code modified to account forhydrofracturing as discussed above. As the excavation of Sub-Cell 1reached its maximum depth (176 mASL), tensile stresses were devel-oped at the interface between the bedrock and Black Till (Figure 4a).These tensile stresses initiated a hydrofracture in the lower Black Tilllayer (Figure 4b). High pore pressures from the bedrock built up inhydrofractured zone and reduced the effective stress at the interfacewith unfractured area of the Black Till. When the effective stress re-duced to zero, the hydrofractures propagated upwards normal tothe minimum principal stress direction (Figure 4c). This model pre-dicted that venting would occur near the center of excavation base(Figure 4d) rather than at the location of Seep A which was closerto the edge of the excavation.

Excavation resulted in a reduction of the total and effective verti-cal stresses beneath the excavation (as would be expected) and someincrease in horizontal total stress in the soil below the excavation be-cause it was a plane strain analysis and the stresses need to find apath to redistribute when the excavation reaches depth. Howeverthe effective stresses were also influenced by the high water pres-sures in the aquifer below the clay layer. For example, the horizontaleffective stresses after excavation of Sub-Cell 1 to an elevation of182 mASL (18 m below OGL; Figure 5a) shows high effective stressesin the middle of the layer and a high stress gradient towards the aqui-fer where stresses were reduced to zero due to high pore pressureexerted by the active aquifer. At this point there is no hydrofracturing.With full excavation of the trench to 176 mASL (24 m below OGL),hydrofractures were initiated at the interface with bedrock and thenpropagated horizontally and then upwards giving an effective stressdistribution as shown in Fig. 5b.

This modeling approach appears to have captured the generalhydrofracturing mechanism that resulted in the venting of gas andwater that was observed but not the precise details, suggesting thatthere may be additional factors that influence the location at whichthe seep developed. This will be explored in following sub-sections.

7.1. Predictions of venting considering hydrofracturing and the presenceof sand lenses

Sand traces were observed with the gas and water venting at SeepA suggesting that the water had passed through a sand lens beforeexiting at the surface of the excavation. Sand lenses are frequently en-countered in the Black Till between elevations 172–176 mASL. Thissection studies the potential effect that both hydrofracturing andsand lenses could have had on the prediction of venting at the loca-tion of Seep A.

Fig. 4. 2D finite element modeling of Sub-Cell 1 excavation using modified code to in-clude hydrofracturing criteria (Section A, Figure 3): (a) hydrofracture initiation at theinterface between the bedrock and Black Till. (b) Hydrofractures spread along interfacebetween bedrock and Black Till. (c) Hydrofractures propagate upwards within theBlack Till. (d) Hydrofractures reach excavated surface.

Fig. 5. Horizontal compressive effective stress field after excavation of Sub-Cell 1 downto a) 182 mASL b) 176 mASL (stresses in kPa).

Fig. 6. 2D modeling the effect of presence of sand lenses within the Black Till onhydrofracturing path (Section A, Figure 3): (a) sand lens intersecting the hydrofracturingpath. (b) Hydrofractures spread across the sand lens.

51R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

7.1.1. Sand lensesTo investigate the possible effect of sand lenses on hydrofracture de-

velopment, the modeling of the unloading sequence was repeated as-suming 10 m long sand lenses at a number of different horizontallocations between elevations 174–175 mASL. In this section the pres-ence of gas in the sand lenses was neglected; it will be considered later.

Since the sand lenses were surrounded by clay of low hydraulicconductivity, lenses initially behave in an undrained mannerduring the analysis. After the excavation depth reached 176 mASL,hydrofractures propagated through the Black Till as discussedabove. If the hydrofracturing path intersected a sand lens on its waytowards excavation surface, (Figure 6a), the fracture transmitted thehigh pore pressure from the bedrock aquifer to the sand lens, reduc-ing the effective stresses within the lens and inducing liquefaction ofthe sand in the lens. As the sand lens liquefied, the hydrofracture con-tinued to propagate to the excavated surface (Figure 6b). Hence, the

analysis suggests that the intersection of a hydrofracture path witha sand lens will accelerate its propagation towards soil surface andwill spread cracks over a larger area of the Black Till.

7.1.2. Gassy sand lensesGassy soils are distinct from unsaturated soils by virtue of the fact

that there are no matric suctions; the pore pressures are positive.Gassy soils may be subdivided into two classes based on the gas bubblesize relative to soil particle size “small” bubbles in coarse grained soil

Fig. 7. 2D modeling the effect of presence of gassy sand lenses within the Black Till onhydrofracturing path (Section A, Figure 3): (a) hydrofracture initiation at interface be-tween the bedrock and Black Till and around gassy sand lenses. (b) Hydrofracturesattracted towards gassy sand lenses. (c) Hydrofractures propagate from sand lens to-wards excavated surface.

52 R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

and “large” bubbles in fine grained soil. For coarse grained soil (such assand lenses), the gas is in the form of bubbles that are smaller than thepores and hence are contained within the pore fluid in the pores.Thus the presence of even a small amount of gas within the pore fluid(e.g., arising from exsolution of gas on unloading) can significantly in-crease pore fluid's compressibility (Brandes, 1999; Sobkowicz andMorgenstern, 1984).

Mabrouk and Rowe (2011) studied the possible role that gassy sandlenses could play as in the absence of hydrofracturing and concludedthat gas in the sand lenses may have contributed to the venting phe-nomenon but only as a secondary factor. This section examines howgassy sand lenses might have contributed to hydrofracturing and gasventing.

The 2D modeling with hydrofracturing was repeated taking accountof the change in pore fluid compressibility due to the presence of smallgas bubbles within sand lenses as described by Mabrouk and Rowe(2011). Briefly, gassy sand lenses can be modeled as a two phasematerial: a solid phase and compressible pore fluid (Harris andSobkowicz, 1977; Dusseault, 1979; Cheung, 1985; Byrne and Jenzen,1984; Vaziri, 1986 and Grozic, 2005). Total stress is distributed betweenthe two phases based on their relative compressibility. Since the gassypore fluid is much more compressible than soil particles, the reductionin total stress due to unloading is shifted to the soil particles causing re-duction in the effective stress in the sand lens due to an almost constantpore pressure (Mabrouk andRowe, 2011). In the present case, during ex-cavation, the gassy sand lenses behaved in a drained manner despitebeing surrounded by a layer of low hydraulic conductivity (Figure 7a).Eventually, as effective stresses are reduced to zero, the sand lens lique-fied causing initiation of minor hydrofractures in the adjacent Black Tillindependent of hydrofracturing from the bedrock. At the same time, ex-cavation of Sub-Cell 1 also caused hydrofracturing starting at the bedrockinterfacewith the Black Till (Figure 7a). These hydrofractures propagatedtowards more compliant zones and kept propagating along a pathwhere the minimum principal stress became tensile (Figure 7b). Thehydrofractures propagated to zones of high pore pressures (lower effec-tive stress) in sand lenses (Figure 7b). The hydrofractures from the bed-rock and sand lenses then merge together towards the excavationsurface (Figure 7c).

Regardless of their connection with the hydrofractures emergingfrom bedrock, sand lenses have the potential of causing minor second-ary venting at some locations due to lens liquefaction during unloading(Mabrouk andRowe, 2011) (Figure 7a,b,c). However, the expected levelof venting of a realistic sand lens size would be insignificant comparedto gas volumes released at the site (Mabrouk and Rowe, 2011). Hence,it can be postulated that the presence of gassy sand lenses – withinclose proximity to hydrofracturing path –might connect hydrofracturesto sand lenses and hence spread cracks over a wider area of the BlackTill. This is consistent with field observations of sand traces in somegas vents (Seep A).

7.1.3. Seep A expected scenarioBased on consideration of the effect of the different variables intro-

duced earlier, a scenario is now developed that can explain the ventingat the actual location of Seep A. The excavation of Sub-Cell 1 wasmodeled assuming the presence of two adjacent 10 m wide×1 mthick sand lenses at elevations 180 and 174 mASL (Figure 8). The sandlenses were given gassy soil properties and were positioned towardsthe southern edge of the sub-cell (close to location of Seep A).

Excavation of Sub-Cell 1 down to elevation 176 mASL initiatedhydrofracturing at the interface of the Black Till with the bedrock(Figure 8a). Simultaneously, the gassy sand lenses (behaving in an essen-tially drainedmanner) sustainedmost of their high pore pressures as ex-solution of gas allowed volume change in the lenses during unloading.This lead to a reduction in effective stress in the sand lenses whicheventually caused the sand to liquefy, producing minor hydrofracturesaround their surfaces (Figure 8a). With time, hydrofractures initiated at

the base of the Black Till propagated upwards towards zones of highpore pressure created by the liquefied sand lenses (Figure 8b). Mean-while, hydrofractures produced by sand lenses spread around their sur-faces, connecting the lenses together forming a larger weak zone withinthe Black Till (Figure 8c). Eventually, the hydrofracturing path emergingfrom the Basal Till intersected with the gassy sand lens which thenspread the hydrofractures around its surface, propagating to the soil sur-face around the location of Seep A (Figure 8d,e). It can be concluded thatexcavation of Sub-Cell 1 down to elevation 176 mASL combinedwith thepresence of gassy sand lenses lead to a continuous seep emerging frombedrock and passing through sand lenses towards the soil surface atthe location where Seep A was observed.

The scenario explains the venting location and appearance of sandtraces in Seep A. The scenario emphasizes the role of the sand lens increating a weak zone within Black Till and directing the path ofhydrofracture propagation if located within close proximity. The sce-nario is supported by the hydrogeological reports that suggested that,based on dissolved minerals, the venting source of the water at SeepA was a mixture of aquifer water and water from the Black Till layer.

8. Sub-Cell 2

The results of the modeling (Figure 9a–e) show that excavatingSub-Cell 2 to its full depth (24 m) caused a significant reduction in

Fig. 9. Hydrofracture formation and propagation after full excavation of Sub-Cell 2 (Sec-tion B, Figure 3): (a) hydrofracture initiation at the interface between the bedrock andBlack Till. (b) Hydrofractures spread across the interface between the bedrock withBlack Till. (c) Hydrofractures propagate upwards within the Black Till. (d) Hydrofracturesreach excavated surface towards southern edge (Seep B SW location). (e) Hydrofracturesspread along excavated base.

Fig. 8. Hydrofracturing scenario for Seep A assuming the presence of gassy sand lensesaround the southern edge of Sub-Cell 1. Hydrofracture formation and propagation afterfull excavation of Sub-Cell 1 is presented (Section A, Figure 3): (a) hydrofracture initiationat the bedrock interface with Black Till and around gassy sand lenses. (b) Hydrofracturesfrom bedrock propagate upwards through Black Till. (c) Hydrofractures from bedrock areattracted towards gassy sand lens. Hydrofractures produced around gassy sand lenses ex-pand and connect between sand lenses. (d) Hydrofractures propagating from bedrock in-tersect with gassy sand lenses. (e) Hydrofractures from bedrock and gassy sand lensesconnect, spread around gassy sand lens surface andpropagate towards soil surface aroundexpected location of Seep A.

53R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

effective stress at the interface between bedrock and Black Till. Thisgave rise to hydrofracturing (Figure 9a). The hydrofractures werepredicted to propagate through the Black Till upwards towards exca-vation base following the path of the minimum principal stress(Figure 9b,c). The analysis indicated that the hydrofractures wouldreach the excavated surface towards the south toe of Sub-Cell 2(close to the location of Seep B SW—Figure 9d). Hence, the model sug-gests that excavation of Sub-Cell 2 to its full excavation depth

(176 mASL) was sufficient to cause the venting observed at SeepB-SW (Figure 9d) without the presence of any sand lenses. No sandwas observed in the venting water at this location. Eventually,hydrofractures were predicted to spread along Sub-Cell 2 base(Figure 9e). This can be linked to the appearance of minor secondaryventing in different locations in Sub-Cell 2. High concentrations of so-dium and chloride in the venting water support the prediction thatthe aquifer is the source of evolving water.

After full excavation of Sub-Cell 2, it was divided into two sectionsby placing a dividing berm of compacted clay up to 182 mASL elevation.Modeling of the loading due to the placement of the sub-cell dividerwas found to direct the hydrofracture flow towards the north ofSub-Cell 2 base; this diversion of the hydrofracturing path explainsthe venting of Seep B-NE (Figure 10).

Fig. 10. Sub-Cell 2 loading after adding a dividing berm causes a detour of thehydrofracturing path towards the excavation of the northern end (Seep B NE) (SectionB, Figure 3).

54 R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

8.1. Predictions of venting considering hydrofracturing and the presenceof intermittent Basal Till

Drilling indicated the presence of a discontinuous weak layer (BasalTill) separating the Black Till and the bedrock in some locations. BasalTill is characterized by its lower shear strength parameters, lower stiff-ness, and relatively high hydraulic conductivity compared to the BlackTill (Tables 1, 2). Rowe and Mabrouk (2007) performed analyseswhich suggested that from considerations of stability, it is unlikelythat this layer is continuous beneath the section of interest. However,while discontinuous, the local presence of Basal Till in some areas mayhave affected the evolution of hydrofractures. To investigate the possi-ble impact of the local presence of the Basal Till on hydrofracturing dur-ing excavation, a localized 15 m wide and 2 m thick layer of Basal Tillwas modeled for different locations.

An analysis was conducted with Basal Till layers of 15 mwidth, 2 mthickness under the edges of Sub-Cell 2. During excavation, it was pre-dicted that the hydrofractures would propagate diagonally from theBasal Till unit (located under the northern edge of the excavation to-wards the southern edge of the Sub-Cell 2) through the Black Tilllayer (Figure 11). Hence it might be postulated that the observation ofthe 45° inclined cracks can be due to the migration of hydrofracturesfrom the point of initiation at the edge of the Basal Till towards the re-mote side of excavation toe following the path of the minimum princi-pal stress. 3D modeling would be needed to validate this scenario andevaluate the effect of discrete Basal Till locations on the hydrofracturingangle and directions.

9. Summary and conclusion

A gas and water venting incident was encountered during excava-tion in a deep clay silt layer. The paper presents a forensic modelingstudy to explain how this venting may have developed.

The site stratigraphy involved 40 m clayey glacial deposits overlyingweathered bedrock that acts as an aquifer and is believed to be thesource of natural gas. Conventional 2Dmodeling of the problemshowedthat local tensile stress concentration at the interface of bedrock and theclayey till develops during excavation. This observation suggested thepotential for the initiation of hydrofracturing originating at the interface

Fig. 11. Hydrofractures propagate diagonally through Black Till in Sub-Cell 2 due to thepresence of discontinuous Basal Till layer under edge of excavation (Section B,Figure 3).

between the bedrock and the till. A FE code wasmodified to account forhydrofracturing.

2Dmodeling of hydrofracturing showed that excavation of Sub-Cell 1to its full depth would initiate hydrofractures at the interface betweenclayey till and bedrock. These hydrofractures would propagate upwardstowards the excavated surface causing gas venting but not at the locationwhere gas venting was observed for Sub-Cell 1 (Seep A). Since thewateremanating from Seep A contained sand traces, it was suspected that theventing path passed through one or more sandy lenses in the clayey till.When appropriately placed sand lenses were included in the model, thehydrofractures emanating from bedrock and passing through the lensesmerged at the location where Seep A was observed.

It is concluded that the venting that was observed at Seep A can beexplained by hydrofracturing from bedrock caused by stress redistribu-tion with the unloading associated with the excavation of Sub-Cell 1combined with the presence of gassy sand lenses near the locationwhere venting was observed. This scenario is consistent with the pres-ence of sand traces in the seep water and the geochemistry of the seepwater that implied amixture of water for the bedrock and the clayey tillsand lenses.

It is also concluded that if these sand lenses had not been present,venting would have still occurred at the time of full excavation to el-evation 176 mASL (24 m depth from the original ground surface), butat a different location.

Modeling of the excavation of Sub-Cell 2 indicated the developmentof Seep B due to hydrofracturing initiated at the bedrock interface withthe clayey till. The modeled hydrofractures were directed towards thesouth-west location where venting was observed. Modeling the subse-quent placement of a divider berm of Sub-Cell 2 predicted that thehydrofracture's path would move towards the north end of the cell towhere the north–east expression of Seep B was actually observed.Again the modeling of hydrofracturing gives results consistent withthe field observations.

The effect of the presence of discontinuous/intermittent weak basaltill layer between bedrock and clayey till was investigated. The modelsuggested that presence of remote discrete basal till layer below theedges of Sub-Cell 2 may be responsible of the hydrofractures propagat-ing diagonally (45°) throughout the clayey till layer although 3Dmodel-ing would be required to confirm this hypothesis.

Acknowledgments

The research reported in this paper was funded by the Natural Sci-ence and Engineering Research Council of Canada (NSERC). The com-putations reported in this paper were performed at the HighPerformance Computing Virtual Laboratory (HPCVL) which is fundedby the Canada Foundation for Innovation, the Government of Ontario,and NSERC. We greatly appreciate the access to data and reports pro-vided by Safety-Kleen Ltd. which, together with Dittrich (2000) pro-vided much of the geological, hydrogeological and geotechnical dataused in developing the idealization examined in the analyses. The au-thors greatly appreciate the value of discussion with G. Funk regard-ing the site hydrogeology. Although we very much appreciate thehelp and assistance of Safety-Kleen Ltd and G. Funk, the authorstake full responsibility for the interpretation and opinions expressedherein. The authors greatly appreciate the constructive comments ofthe anonymous reviewers.

References

Beissel, S., Johnson, G., Popelar, C., 1998. An element-failure algorithm for dynamic crackpropagation in general directions. Engineering Fracture Mechanics 61, 407–426.

Belytschko, T., Black, T., 1999. Elastic crack growth in finite elements with minimalremeshing. International Journal for Numerical Methods in Engineering 45,601–620.

Belytschko, T., Krongauz, Y., Organ, D., Fleming, M., Krysl, P., 1996. Meshless methods:an overview and recent developments. Computer Methods in Applied Mechanicsand Engineering 139, 3–47.

55R.K. Rowe, A. Mabrouk / Engineering Geology 151 (2012) 47–55

Bjerrum, L., Nash, J.K.T.L., Kennard, R.M., Gibson, R.E., 1972. Hydraulic fracturing in fieldpermeability testing. Geotechnique 22 (2), 319–332.

Brandes, H.G., 1999. Predicted and measured geotechnical properties of gas-chargedsediments. International Journal of Offshore and Polar Engineering 9 (3), 219–225.

Brenner, S.L., Gudmundsson, A., 2004. Arrest and aperture variation of hydrofracturesin layered reservoirs. In: Cosgrove, J.W., Engelder, T. (Eds.), The Initiation, Propaga-tion and Arrest of Joints and Other Fractures: Geological Society of London, SpecialPublication, 231, pp. 117–128.

Byrne, P.M., Jenzen, W., 1984. INCOIL. Soil Mechanics Series No. 80. Department of CivilEngineering, University of British Columbia, Vancouver, Canada.

Chang, H., 2004. Hydraulic Fracturing in Particulate Materials. Ph.D. thesis, Georgia In-stitute of Technology. (276 pages).

Cheung, K.F.H., 1985. A stress–strain model for the undrained response of oil sand.Ph.D. thesis, University of British Columbia, Vancouver, Canada. (137 pages).

Dittrich, J.P., 2000. Slope behavior during excavation of the Sarnia approach to the St Clairtunnel. Ph.D. Thesis, University of Western Ontario, London, Ontario (795 pages).

Dittrich, J.P., Rowe, R.K., Becker, D.E., Lo, K.Y., 2010. Influence of ex-solved gases onslope performance at the Sarnia approach cut to the St. Clair Tunnel. Canadian Geo-technical Journal 47 (9), 971–984.

Dusseault, M.B., 1979. Undrained volume and stress change behavior of unsaturatedvery dense sands. Canadian Geotechnical Journal 16 (4), 627–640.

Grozic, J.L.H., 2005. On the undrained shear strength of gassy clays. Computers andGeotechnics 32 (7), 483–490.

Harris, M.C., Sobkowicz, J.C., 1977. Engineering behavior of oil sand. In: Redford, D.A.,Winestock, A.G. (Eds.), The Oil Sands of Canada–Venezuela: Can. Inst. Min. Metal-lurgy. Special, 17, pp. 270–281.

Ingraffea, A.R., 2008. Computational Fracture Mechanics, In: Stein, E., de Borst, R.,Hughes, T. (Eds.), 2nd Edition. Encyclopedia of Computational Mechanics, volume2. John Wiley and Sons. Chapter 11.

Jaworski, G.W., Duncan, J.M., Seed, H.B., 1981. Laboratory study of hydraulic fracturing.ASCE Journal of Geotechnical Engineering 107 (GT6), 713–733.

Lo, K.Y., Kaniaru, K., 1990. Hydraulic fracture in earth and rock-fill dams. CanadaGeotechiques 27, 496–506.

Mabrouk, A., Rowe, R.K., 2011. Effect of gassy sand lenses on a deep excavation in aclayey soil. Engineering Geology 122, 292–302.

Martha, L., Gray, L., Ingraffea, A., 1992. Three-dimensional fracture simulation with asingle-domain, direct boundary element formulation. International Journal ofNumerical Methods in Biomedical Engineering 35, 1907–1921.

Martini, A.M., Walter, L.M., Budai, J.M., Ku, T.C.W., Kaiser, C.J., Schoell, M., 1998. Geneticand temporal relations between formation waters and biogenic methane: UpperDevonian Antrium Shale, Michigan Basin, USA. Geochimica et Cosmochimica Acta62, 1699–1720.

Massarsch, K.R., 1978. New aspects of soil fracturing in silt-clay. Journal of the Geotech-nical Engineering Division, Proceedings of ASCE 104 (GT8), 1,109–1,123.

Murdoch, L.C., 1988. Innovative delivery and recovery using hydraulic fracturing. Proc.,14th Annual USEPA Research Symposium, p. 613.

Murdoch, L.C., 1989. A field test of hydraulically fracturing glacial till. Proc., 15th AnnualUSEPA Research Symposium, Cincinnati.

Murdoch, L.C., 1993a. Hydraulic fracturing of soil during laboratory experiments, part I,methods and observations. Geotechnique 43 (2), 255–265.

Murdoch, L.C., 1993b. Hydraulic fracturing of soil during laboratory experiments, partII, propagations. Geotechnique 43 (2), 267–276.

Murdoch, L.C., 1993c. Hydraulic fracturing of soil during laboratory experiments, partIII, theoretical analysis. Geotechnique 43 (2), 277–287.

Murdoch, L.C., 1995. Forms of hydraulic fractures created during a field test in fine-grained glacial drift. The Quarterly Journal of Engineering Geology 28, 23–35.

Murdoch, L.C., 2000. Remediation of organic chemicals in the vadose zone. In: Falta, R.,Looney, B. (Eds.), Vadose Zone, Science and Technology Solutions. Battelle Press,pp. 948–1247 (Chapter 7).

Murdoch, L.C., Slack, W.W., 2002. Forms of hydraulic fractures in shallow fine-grainedformations. Journal of Geotechnical and Geoenvironmental Engineering 128 (6),479–487.

Ortiz, M., Pandolfi, A., 1999. Finite-deformation irreversible cohesive elements forthree-dimensional crack propagation analysis. International Journal for NumericalMethods in Engineering 44, 1267–1282.

Panah, A.K., Yanagisawa, E., 1989. Laboratory studies on hydraulic fracturing criteria insoil. Soils and foundations. Japanese Society of Soil Mechanics and Foundation En-gineering 29 (4), 14–22.

Quigley, R.M., Ogunbadejo, T.A., 1976. Till geology, mineralogy and geotechnical behav-ior, Sarnia, Ontario. In: Legget, R.F. (Ed.), Glacial Till, An Inter-disciplinary Study:Royal Society of Canada, Special Publication, 12, pp. 336–345.

Rowe, R.K., Mabrouk, A.B., 2007. Three dimensional effects of excavation on stress dis-tribution. 60th Canadian Geotechnical Conference, Ottawa, October, pp.1142–1147.

Sobkowicz, J.C., Morgenstern, N.R., 1984. The undrained equilibrium behavior of gassysediments. Canadian Geotechnical Journal 21, 439–448.

Sture, S., Alqasabi, A., Ayari, M., 1999. Fracture and size effect characters of cementedsand. International Journal of Fracture 95, 405–433.

Van der Pluijm, B., Marshak, S., 2004. Earth Structure: An Introduction to Structural Geol-ogy and Tectonics. WW Norton & Company . Chapter 7.

Vaziri, H.H., 1986. Nonlinear temperature and consolidation analysis of gassy soils.Ph.D. thesis, Department of Civil Engineering, University of British Columbia, Van-couver, Canada.

Wang, L., Brust, F.W., Atluri, S.N., 1997. The elastic-plastic finite element alternatingmethod (EPFEAM) and the prediction of fracture under WFD conditions in aircraftstructures part I: EPFEAM theory. Computational Mechanics 19, 356–369.

Wang, J., Zhu, J., Chiu, C.F., Zhang, H., 2007. Experimental study on fracture toughnessand tensile strength of a clay. Engineering Geology 94, 65–75 (2007).

Weihe, S., Kroplin, B., De Borst, R., 1998. Classification of smeared crack models basedonmaterial and structural properties. International Journal of Solids and Structures35, 1289–1308.

Zhu, X., Joyce, J.A., 2012. Review of fracture toughness (G, K, J, CTOD, CTOA) testing andstandardization. Engineering Fracture Mechanics 85, 1–46.