Algorithm for generation of stratigraphic profiles using cone penetration test data

Computers and Geotechnics 90 (2017) 73–84

Contents lists available at ScienceDirect

Computers and Geotechnics journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Algorithm for generation of stratigraphic profiles using cone penetration test data Eshan Ganju ⇑, Monica Prezzi, Rodrigo Salgado Purdue University, Lyles School of Civil Engineering, 550 W. Stadium Avenue, West Lafayette, IN 47906, USA

a r t i c l e

i n f o

Article history: Received 13 September 2016 Received in revised form 11 April 2017 Accepted 26 April 2017

Keywords: Cone penetration test Soil behavior type charts Stratigraphic profile Thin layers

a b s t r a c t Cone Penetration Test (CPT) data are often used directly in the design of shallow and deep foundations and many other applications. To produce more cost-effective designs, it is advantageous to use CPT data to establish stratigraphic profiles as well. Algorithms to generate a stratigraphic profile using data from an individual CPT sounding and a Soil Behavior Type (SBT) chart as inputs are presented. Two SBT charts from the literature were selected and modified to eliminate ambiguity in soil classification. Novel algorithms were developed for handling the occurrence of thin layers within a stratigraphic profile to account for the fact that the standard CPT cone cannot accurately sense layers with thickness below a certain limit and a representative cone resistance cannot be obtained if the layer is too thin. Likewise, the algorithms prevent the creation of a soil profile with adjacent layers of essentially the same soil by consolidating layers appropriately. The algorithms presented generate a design soil profile, produced using a precise classification based on soil type and state and by elimination of artificial layering, that can be more effectively used in design. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Over the years, the cone penetration test (CPT) has gained acceptance as a fast, reliable and economical tool for soil profile characterization [4,16,17,22,33,36,39,42,50,51,53,56], foundation design [5,7,13,14,15,19,18,20,21,25,26,30,29,28,31,35,59] and liquefaction susceptibility [8,11,38,40,44,46]. One of the primary applications of the cone penetration test is stratigraphic profiling [6,12,32,39,41,57,60]. Stratigraphic profiling can be most useful in design if soil layers or strata consist of a well-defined soil type with equally well-defined state (typically density, indirectly represented by the degree of overconsolidation in clayey soils). Once clarity exists about the soil in an individual layer, then it is reasonable to compute statistics (expected value/trends, coefficient of variation, scale of fluctuation) for such a layer. The quantification of soil profile variability can lead to better choice of resistance factors in the design of footings, piles, slopes and embankments when following the LRFD method [1,19,18,23,24,34,47,54]. Typically, soil samples are not collected at the exact location where a CPT sounding is performed, so soil profiles are obtained from SPT tests performed nearby or inferred from the CPT test data using soil behavior type (SBT) charts. An SBT chart serves as a sim⇑ Corresponding author. E-mail address: [email protected] (E. Ganju). http://dx.doi.org/10.1016/j.compgeo.2017.04.010 0266-352X/Ó 2017 Elsevier Ltd. All rights reserved.

ple signal transfer function that converts cone resistance-skin friction pairs to ‘‘soil behavior” types. In using the CPT for any type of interpretation, it is important to keep certain core ideas in mind. For example, it is important to distinguish between soil intrinsic and state variables [3,46,48]. Intrinsic variables do not depend on the state of the soil, as defined by density, stress state and structure. Intrinsic variables would most closely relate to soil composition (in terms of the usual particle size-based terms such as clay, silt or sand). The measurements made by a CPT reflect both state and intrinsic variables, and so reflect the soil constitutive response (or ‘‘behavior”) in its totality. Hence, layering should be defined in terms of both composition (e.g., ‘‘sand”) and state (e.g., ‘‘dense”) with accompanying quantitative data as the core of what is obtained from the CPT. Additionally, independent sampling to define composition and even intrinsic and state variables would be superior to interpreted values from a CPT test if possible to perform reliably. As a result of the nature of the results obtained from the cone penetration test, the generation of soil profiles from SBT charts is subject to a degree of uncertainty. Robertson [42] gave examples of how soil behavior types obtained from SBT charts need not be in agreement with the traditional soil classifications based on grain-size distribution and soil plasticity (e.g., USCS soil classification). According to Robertson [42], differences in soil classification are likely to result from compositional and behavioral points of

74

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

view. This is particularly true for mixed soils, for which the nature of the soil fabric plays a major role [9,10,42,45]. This paper presents the framework of a comprehensive algorithm for analysis of CPT test data for obtaining soil behaviorbased stratigraphic profiles. These profiles are developed with the aim of understanding the nature of the behavior of the soil tested and to group regions of similar soil behavior within a soil profile appropriately. The paper has been divided into three main sections. The first section details the two modified SBT charts developed in this research, the second section focuses on the algorithm developed for generation of soil profiles using CPT data, and the third section gives an example of a soil profile generated using the algorithm proposed in this paper. The applicability of the algorithm is general, and thus the algorithm can be implemented for any SBT chart available in literature. 2. Modified soil behavior type charts

(a) Cla yey Sil ty San d

500

Clean Sand or Silty Sand

Sa

nd

or

Si

lt

100

ay

ey

Very Stiff Clay

Cl

Cone resistance qc/pA

Many soil behavior type (SBT) classification charts have been proposed over the years. Some of the early SBT charts are those of Begemann [4], Sanglerat et al. [50], Schmertmann [51], Douglas and Olsen [16], Tumay [56], Robertson et al. [43], Senneset et al. [55], Robertson [39], Larsson and Mulabdic [27], and Jefferies and Davis [22]. Some of the more recent SBT charts include those by Ramsey [36], Schneider et al. [52,53], and Robertson [37,42]. While the soil profile generation algorithm that will be described in the subsequent sections can be used for any SBT chart in the literature, modified versions of the SBT charts by Tumay [56] and Robertson [39] were used to generate the soil profiles in this paper. This choice was made partly because of familiarity of engineers with these two charts and partly because the charts had certain core features that are required for a logical, quantitative algorithm to be used in CPT interpretation. Modifications to the selected charts were made to minimize ambiguities associated with behavior types. These modifications were necessary to develop a logical stratigraphic profiling algorithm using CPT data.

Sandy Clay or Silty Clay

(1) The regions of the original chart (‘‘loose sand”, ‘‘sand”, ‘‘shell sand or limerock”, ‘‘dense or cemented sand” and ‘‘silty sand”) were removed and consolidated into a single region referred to as ‘‘clean sand or silty sand”. When a soil falls into this ‘‘clean sand or silty sand” region of the chart, it is further classified into five different subtypes depending on the estimated relative density (from very loose to very dense), as shown in Table 1. Primarily, these subtypes serve to describe the state of the soil in situ. The relative density of the sandy soil is calculated using CPT data, as suggested by Salgado and Prezzi [49]. (2) The ‘‘silty clay” region in the original chart was removed, as it was believed that the resolution of the CPT may not be enough to distinguish, with a sufficient degree of certainty, ‘‘silty clay” from ‘‘sandy clay”. (3) The ‘‘sandy clay” region in the original chart was renamed ‘‘sandy clay or silty clay”. (4) The ‘‘organic clay” region in the original chart was also removed because the resolution offered by qc-fs pairs may not be enough to distinguish ‘‘organic clay” from the neighbouring ‘‘inorganic clay”. The ‘‘inorganic clay” regions of different stiffnesses in the original chart were changed to ‘‘clay” of different stiffnesses in the modified chart.

y

Medium Stiff

Clayey Silt

2.1. Modified Tumay chart Fig. 1(a) and (b) shows the original and modified Tumay charts. The following modifications were made to the original Tumay chart [56]:

Stiff Cla

10

Clay

Soft Clay

Very Soft Clay Sensitive Clay

1 0

1

2

3

4

5

6

Friction ratio (%)

(b) Fig. 1. SBT charts: (a) original Tumay chart [56] and (b) modified Tumay chart.

Table 1 Sand classification according to density. Relative density (%)

Sand classification

0–15 15–35 35–65 65–85 85–100

Very loose sand Loose sand Medium dense sand Dense sand Very dense sand

(5) The ‘‘clayey sands” region in the original chart was changed to ‘‘clayey silt” in the modified chart. This was done to be consistent with the expected progressive increase in cone resistance with increasing sand content from ‘‘clayey silt” to ‘‘clayey sand or silt” and then to ‘‘clayey silty sand”. (6) A new region, ‘‘sensitive clay”, was added. This region in the modified SBT corresponds to clays with FR less than unity, which is often found to suggest the presence of sensitive clays [17].

75

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

2.2. Modified Robertson chart The original Robertson chart [39] uses normalized cone resistance and normalized friction ratio values. To do the normalization, additional information on the unit weights of the soils found at the test location and the ground water elevation are required. These charts can only be used after post-processing of the CPT data. The normalized charts by Robertson [39] are expected to be more reliable than the non-normalized charts proposed by Robertson et al. [43]. However, when the in situ vertical effective stress is between 50 kPa and 150 kPa, there is often little difference between the resulting normalized and non-normalized soil behavior types [42]. Fig. 2(a) and (b) shows the original and modified Robertson charts, respectively. The following modifications were made to the original Robertson chart [39]:

1. The ‘‘organic soils – peats” region (zone 2) was renamed ‘‘organic clay” in the modified chart. 2. The ‘‘very stiff sand to clayey sand” region (zone 8) in the original chart was incorporated into two different regions: ‘‘sand mixtures: silty sand to sandy silt” (zone 5 of the original chart) and ‘‘clean sand to silty sand” (zone 6 of the original chart). The boundary line between zones 5 and 6 was extended all the way to the top axis (the normalized friction ratio axis) in the modified chart. The rationale behind this modification is that the ‘‘very stiff sand to clayey sand” region (zone 8) in the original chart is similar to the ‘‘sand mixtures: silty sand to sandy silt” region (zone 5) and ‘‘clean sand to silty sand” (zone 6) regions. 3. The ‘‘clean sand to silty sand” region (zone 6) was further classified into five different subtypes depending on the estimated relative density (very loose to very dense). The relative density of the sandy soils was calculated using CPT data, as suggested by Salgado and Prezzi [49].

(a) 1000

Normalized cone resistance

Gravelly Sand to Sand

100

Clean Sand to Silty Sand

: res ilt xtu dy S i M n a d S n Sa d to ay n Cl Sa y ilty t l i S S t to Sil ay l C

10

Clay to Silty Clay Sensitive Fine Grained Organic Clay 1 0.1

1

10

Normalized friction ratio

(b) Fig. 2. SBT charts: (a) original Robertson chart [39] and (b) modified Robertson chart.

76

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

Fig. 3. Soil profile generation flowchart – part 1.

4. The ‘‘very stiff, fine-grained” region (zone 9) in the original chart was incorporated into the ‘‘clay to silty clay” (zone 3 of the original chart) and ‘‘clayey silt to silty clay” (zone 4 of the original chart) regions in the modified chart. The boundary line between zones 3 and 4 was extended all the way to the top axis (the normalized friction ratio axis) in the modified chart. The rationale behind this modification is that the ‘‘very stiff, finegrained” (zone 9) region in the original chart indicates similar composition to that of ‘‘clay to silty clay” (zone 3) and ‘‘clayey silt to silty clay” (zone 4) regions.

the initial soil profile are absorbed into thick layers by the soil profile generation algorithm according to three different approaches: 1. SBT chart band approach (consolidation of thin layers into adjacent layers considering secondary soil type(s) classification). 2. Soil group approach (consolidation of thin layers into adjacent layers of the same soil group). 3. Average qc approach (consolidation of thin layers into adjacent layers with similar average qc). 3.1. Initial soil profile

These modifications were needed to have a clearer distinction between soil intrinsic variables (related closely to composition) and soil state variables. It becomes impossible, as an example, to quantify soil variability, which is closely linked to soil composition (as in, for example, ‘‘clay” versus ‘‘sand”) when what could be viewed as a sand appears in more than one zone. 3. Soil profile generation algorithm Using the modified soil behavior charts as signal transfer functions, the CPT data can be analyzed to obtain the soil behavior profiles at a site. The procedure consists of two main steps (as shown in Figs. 3 and 4): (1) obtain an initial soil profile by ‘‘plotting” the CPT data on the modified SBT charts; (2) modify the initial soil profile by consolidating thin layers, to obtain the final soil profile. After obtaining the initial soil profile, the first step of the procedure is to classify soil layers as either thick or thin. All thin layers in

The initial soil profile follows straight from the CPT data (qc, fs, FR, pore pressure and possibly other quantities versus depth). The FR and qc values obtained at each depth are first plotted on the selected SBT chart. Then a ‘‘primary soil behavior type” is assigned to each depth depending on where the point falls on the SBT chart selected. This process is repeated for all depths of the CPT sounding for which data was obtained. Subsequently, all adjacent depths with the same SBT classification are grouped into initial soil layers. All the initial soil layers put together define the initial soil profile. 3.2. Thin layer consolidation The initial soil profile consists of layers of various thicknesses. Some layers can be quite thin (e.g., only a few centimeters thick). However, since the standard cone has a diameter of 35.7 mm, it has a minimum resolution for layer identification. The reason for

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

77

distance [2]. In addition, as the cone approaches a layer of different stiffness, it will start sensing its presence a few cone diameters ahead of the interface between the layers [2,58]. These development and sensing distances depend on the relative stiffness of the layers traversed by the cone. If both the sensing and development distances are considered, then layers thinner than approximately 150–200 mm cannot be properly detected by the standard CPT cone, which can result in ambiguity in the assignment of soil type to such layers. Accordingly, this research considers a layer having a thickness of 150 mm or less (corresponding to about 4.2 standard cone diameters) a thin layer.

Fig. 4. Soil profile generation flowchart – part 2.

this limitation is that the cone needs to penetrate a certain depth into a given layer to develop the cone resistance that is representative of that layer. This distance is referred to as the development

3.2.1. Merging of thin layers by SBT band approach Bands were introduced in the modified charts to address the uncertainty associated with the location of the boundary lines between regions defining different soil composition and state in a chart. When generating a soil profile, it is possible that a thin layer, as defined previously, may be produced by CPT data plotting very close to a boundary line between two regions in the chart. A thin layer produced in this manner can be argued to be a part of a layer adjacent to it in the soil profile. In such cases, thin layers may be consolidated into adjacent thick or thin layers if they are sufficiently similar. It must be noted here that a thick soil layer (thickness greater than 150 mm) need not be consolidated into adjacent layers if it lies close to the boundaries on the chart. The reason being that thick layers allow for sufficient penetration of the CPT probe to develop the shaft and base resistance representative of the layer. Fig. 5 shows an idealized segment of a soil profile where we can see two layers: one thin layer of type ‘‘clean sand or silty sand” and one thick layer of type ‘‘clayey sand or silt”. Assume that the thin layer of ‘‘clean sand or silty sand” were produced by CPT data falling close to the boundary line between these two soil behavior types in the modified Tumay chart, so that it could also be assigned a secondary classification of ‘‘clayey sand or silt” based on the proximity of the cone resistance-friction ratio pair to the boundary between ‘‘clean sand or silty sand” and ‘‘clayey sand or silt”, as shown in Fig. 5. The thin layer could then be consolidated into the adjacent ‘‘clayey sand or silt” layer since the secondary soil behavior classification of the thin layer matches the primary soil behavior classification of the thick layer. To be able to assign a secondary soil classification to a layer, upper and lower boundary lines corresponding to qc differing by ±15% with respect to the existing boundary lines between soil behavior types were added to the modified SBT charts. These upper

Fig. 5. Thin layer generated by CPT data falling close to the boundary between ‘‘clean sand or silty sand” and ‘‘clayey sand or silt” in the modified Tumay chart.

78

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

Cla yey S

ilty S

and

500

Clean Sand or Silty Sand

nd

or

Si

lt

+ 15% qc bands

ay

ey

Sa

Very Stiff Clay

Cl

Cone resistance qc/pA

100

Sandy Clay or Silty Clay

Stiff Cla y

10

Medium Stiff

Clayey Silt + 15% Friction ratio bands

Clay

Soft Clay

Very Soft Clay Sensitive Clay

1 0

1

2

3

4

5

6

Friction ratio (%) Fig. 6. Modified Tumay chart with ±15% qc and friction ratio bands.

and lower lines define bands in the SBT charts. Fig. 6 shows the modified version of Tumay chart with ±15% qc dashed lines defining these bands (similar bands were added to the modified Robertson chart). There are two vertical lines in the modified Tumay chart (between ‘‘clayey silt” and ‘‘clayey sand or silt” regions and between ‘‘sensitive clay” and ‘‘very soft clay” regions); in these two cases, bands were created by adding ±15% of the friction ratio to the boundary lines. The choice of ±15% was made on the basis that choosing a larger percentage of qc (or friction ratio) to form bands on the modified charts results in the band area becoming too large to be practical.

On the other hand, choosing a smaller fraction of qc results in the bands becoming too small for the intended purpose of quantifying proximity to a boundary in soil ‘‘behavior” type. This number is necessarily subjective and must be selected based on what at the end works in practice, in terms of producing final soil stratigraphy. The stress is placed on the concept rather than on the specific thickness of these bands around boundaries between soil types. The bands represent, in effect, the uncertainty that exists in soil behavior type classification. Whenever a thin layer occurs, the soil profile generation algorithm checks whether it has a qc and FR combination (qc and FR are averaged over the layer thickness) that falls within these bands. If it does, a secondary classification is assigned to it. If the assigned secondary classification of the thin layer matches the primary classification of the neighboring thick layer in the initial soil profile, then the thin layer can potentially be incorporated into the thick layer, pending further checks. There is a possibility that the average qc and FR combination of a thin layer falls at the intersection of two bands. In such a case, the thin layer is assigned two secondary soil classifications. Figs. 7 and 8 show the algorithms developed for merging of thin layers using the SBT band approach. As shown in the flow charts, first the thin layers identified in the initial soil profile are assigned secondary soil types whenever the points with average FR and qc values fall within the bands of the modified chart selected. Secondly, the cone resistance values and soil types are compared to decide whether the thin layer can be incorporated into the adjacent layers or not. A thin layer is incorporated into the adjacent layer if and only if the average qc of the thin layer is within ±25% of the average qc of the adjacent layer. It should be noted here that the ±25% range is different from the ±15% range used to mark bands in the modified charts. The purpose of the ±25% qc range is to prevent the merging of dissimilar layers that may have matching SBTs, while the purpose of the ±15% qc is to form bands around the

Fig. 7. Assignment of secondary soil type to thin layers.

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

79

Fig. 8. Soil profile checked for SBT chart band approach.

Fig. 9. Illustration of the proximity ratio.

Table 2 Soil group classification for the modified Tumay chart. Soil groups

Soil types

Sand Clay

Clean sand or silty sand (very loose to very dense) Sensitive clay, Very soft clay, Soft clay, Medium stiff clay, Stiff clay, Very stiff clay Clayey silt, Clayey sand or silt, Clayey silty sand, Sandy clay or Silty clay

Mixed soil

Table 3 Soil group classification for the modified Robertson chart. Soil group

Soil types

Sand

Gravelly sand to sand (very loose to very dense), Clean sand to silty sand (very loose to very dense) Organic clay, Sensitive fine grained Sand mixtures: Silty sand to sandy silt, Clayey silt to silty clay, Clay to silty clay

Clay Mixed soil

boundaries to account for uncertainty in identification of thin layers which cannot be accurately sensed by the standard cone. Thin layers can occur in various combinations within a soil profile. To develop a comprehensive soil profile generation algorithm, in addition to the algorithms presented in Figs. 7 and 8, it is important that specific scenarios be accounted for and clear rules be estab-

lished to handle the merging of thin layers. The four main scenarios and accompanying rules to handle thin layers are described below: (1) Thin layer adjacent to thick layer: If (a) the secondary soil classification of the thin layer matches the primary soil classification of the thick layer and (b) if the average qc of the thin layer is within ±25% of the average qc of the thick layer, then the thin layer is merged into the thick layer. The newly formed layer is assigned the same classification as that of the thick layer. (2) Thin layer adjacent to another thin layer: Merging of the two layers is attempted if and only if (a) the secondary soil type of one of the thin layers (say, layer A) matches the primary soil type of the other thin layer (say, layer B) and (b) if the average qc values of both layers are within ±25% of each other. If the merging is done, the resulting layer is assigned the primary soil type of layer B. In case of cross matching, i.e., the secondary classification of one matches the primary classification of the other and vice versa, the primary soil type of the resulting layer will be that of the thicker of the two layers. (3) Thin layer sandwiched between two layers: In case the secondary soil classifications assigned to the thin layer under consideration match the primary soil classification of both the layers adjacent to it, the average qc values of the adjacent layers are compared with the average qc value of the thin layer under consideration. Merging of the thin layer is

80

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

Depth (m)

Fig. 10. Soil profile checked for remaining thin layers using the average qc approach.

0

0

0

1

1

1

2

2

2

3

3

3

4

4

4

5

5

5

6

6

6

7

7

7

8

8

8

9

9

9

10

10

10

11

11

11

12

12

12

13

13

13

14

14

14

15

0 10 20 30 40 50 60 Cone resistance (MPa)

15

0

200 400 600 800 Shaft resistance (kPa)

15

0 1 2 3 4 5 6 7 8 9 10 Friction ratio (%)

Fig. 11. Cone penetration test data for test site.

attempted with the adjacent layer for which the average qc value is closer to that of the thin layer under consideration. (4) Cluster of thin layers: Whenever thin layers appear clustered together, the first thin layer to be considered when using the SBT chart band approach is the one whose average (FR, qc) point falls nearest to any of the boundary lines of the modified SBT chart. To decide which of the thin layers in a cluster of thin layers is considered first, the proximity ratio (shown in Fig. 9) for each thin layer in the cluster is calculated. The proximity ratio is defined as the ratio of the absolute value of the difference in cone resistance values between the point representing the thin layer (point C in Fig. 9) and the boundary line

(point B in Fig. 9) to the difference in cone resistance values between the band line (point A in Fig. 9) and the boundary line (point B in Fig. 9) of the modified SBT chart. A very low proximity ratio indicates that the average point representing the thin layer is very close to the SBT chart boundary line; thin layer consolidation progresses in increasing order of proximity ratio. After the initial soil profile is modified to consolidate thin layers using the band approach, the resultant soil profile may have adjacent layers of the same soil type. The soil profile is checked for the occurrence of adjacent layers of the same soil type, and, whenever soil layers of the same soil type occur in sequence, these are consolidated and the soil profile is updated.

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

81

approach is used to further consolidate these remaining thin layers. In the soil group approach, thin layers are consolidated when they belong to the same soil group. Tables 2 and 3 show the soil groups for the modified Tumay chart and the modified Robertson chart, respectively. Three broad groups have been defined based on similarity of behavior of the soils, i.e., sand-like behavior, clay-like behavior and an intermediate category for mixed soils. When thin layers appear in a cluster, the process starts with the thinnest layer in the sequence. The consolidated layer always inherits the soil type of the thicker of the two layers consolidated. After consolidation, the soil profile is checked for the occurrence of adjacent layers of the same soil type, which are merged and the soil profile is updated. 3.2.3. Merging of thin layers by the average qc approach The average qc approach is the last approach used to consolidate the remaining thin layers. This approach consists of consolidating thin layers with the adjacent layer having the closest average qc value. In the case of a cluster of thin layers, the modification process starts with the thinnest layer in the sequence. If a thin layer is at the very top or at the very bottom of the soil profile, it is discarded from the soil profile (with the top elevation of the top layer or bottom elevation of the bottom layer accordingly recorded). Fig. 10 shows the algorithm used to consolidate thin layers using the average qc approach.

Fig. 12. Soil profile obtained from bore-hole logs.

3.2.2. Merging of thin layers by soil group approach After the soil profile is modified using the SBT chart band approach and the similar soil layers appearing in sequence are consolidated, thin layers may remain in the soil profile. The soil group

4. Generated soil profile comparison In order to illustrate the methodology outlined above, soil profiles are generated from standard cone penetration test data collected from a test site located in West Lafayette, Indiana. In this section, the soil profile generated using the proposed algorithm and the modified Tumay chart is compared to (1) the in situ layer

Fig. 13. Soil profile generated by a simple signal transfer function chart (original Tumay chart [56]).

82

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

information obtained from soil borings located near the CPT soundings and (2) the soil profile generated using the original Tumay chart [56] as a simple signal transfer function. While it is known that the SBT-based behavioral classification cannot be compared to the compositional classification obtained from laboratory tests on a one-to-one basis, a qualitative comparison of soil profiles from the three different approaches is nonetheless instructive. Fig. 11 shows the CPT data obtained from the test site in West Lafayette, Indiana, located on the bank of the river Wabash. The CPT sounding was carried out to a total depth of 15 m (49 ft) from the ground surface (surface elevation = 520 ft) with a qc-fs pair collected every 5 cm along the depth. In addition to the CPT sounding, a Standard Penetration Test (SPT) was also carried out near the CPT sounding. The stratigraphic profile obtained from the bore hole logs of the SPT is shown in Fig. 12. Split-spoon samples were collected every 1.5 m (5 ft) to visually assess the type of material. The data from the bore hole logs indicates that the soil profile at the site consists mostly of loose sandy-clay loam on the upper half of the profile with a more medium-dense sandy type of material on the lower half of the soil profile. Gravels and cobbles were also observed on the lower half of the profile during the SPT. In addition to the layering information, Fig. 12 also shows the location of the split spoon sampler and the SPT – N values (observed to range from 2 to 32) obtained during the test at different depths. The data collected from the CPT sounding was used to obtain the soil profile using the original Tumay chart [56], presented in Fig. 1(a), as a simple signal transfer function. Each qc-fs pair obtained from the CPT data was plotted on the original Tumay chart [56] giving the initial point-soil profile. Adjacent data points with the same classification were merged together to generate the soil profile presented in Fig. 13. The profile in Fig. 13 shows a cluster of sandy-clay, clayey-sand type of material in the first third of the profile, with thicknesses ranging from 5 cm to 45 cm. The remaining part of the profile consisting mainly of clusters of dense sands. The profile is littered with clusters of layers, too small to be detected by the CPT cone, resulting in a significantly fragmented profile. Data from the CPT sounding was also processed using the proposed algorithm and the modified Tumay chart. The soil profile generated using the algorithm is shown in Fig. 14. From the profile, two main observations can be made: (1) the generated profile consists of thick layers, with a minimum thickness of 25 cm, well above the sensing and development length of the standard CPT cone, and (2) layers belonging to the clean sand and silty sand region were further classified on the basis of their relative density [49] using Table 1. This allows for a more coherent profile generation and also for separation of the intrinsic and state variablebased descriptors of the in situ conditions. The profile presented in Fig. 14 indicates that the soil profile at the site consists mainly of loose-to-dense silty sand on the upper half with dense sand or silty sand on the lower half, which, in general, agrees with the soil profile obtained from the SPT bore hole logs. It should be emphasized here that the profile generated using SPT bore-hole data is based on soil samples collected at an interval of 1.5 m (5 feet) along the depth of the profile, while the soil profile generated using the algorithm is based on nearly continuous standard cone penetrometer measurements (every 5 cm). Therefore, a more precise one-on-one comparison between the two cannot reasonably be made.

5. Conclusions In this paper, algorithms are presented for the development of stratigraphic profiles using CPT data. First, an initial soil profile is generated by plotting the qc – friction ratio pairs on a given SBT

Fig. 14. Soil profile generated using proposed algorithm and modified Tumay chart.

chart. After an initial soil profile is generated (each soil layer in the profile is assigned a primary soil behavior type classification), layers occurring in the soil profile with thickness less than or equal to 15 cm are tagged as thin layers, layers inside which the CPT probe is unable to develop a cone resistance representative of the soil in the state it exists in that layer. After identification of the initial soil profile and its thin layers, the soil profile is reanalyzed with the goal of merging thin layers into adjacent layers using three main approaches: (1) the SBT band approach, (2) the soil group approach, and (3) the average qc approach. The band approach accounts for the uncertainty associated with the cone resistance-friction ratio pairs of soil layers falling near the boundaries of the SBT regions on the charts. This uncertainty is dealt with by introducing bands (corresponding to ±15% qc or friction ratio) into the modified SBT charts. A secondary soil behavior type is assigned to thin layers that lie within these bands. Different scenarios in which thin layers may occur in the initial soil profile (e.g., in clusters, sandwiched between layers and adjacent to other thin layers) are handled by considering layer thickness, qc similarity and soil behavior type. After consolidation of thin layers using the SBT band approach, thin layers may still exist in the stratigraphic profile. Then the profile is checked using the soil group approach. In the soil group approach, soil behavior types are grouped into soil groups of similar behavior (for example, all soil types that are essentially sand-like are grouped into the sand group and all clay-like soils into a clay group). Thin soil layers are then merged into thick soil layers if they belong to the same soil group. Any remaining ‘‘thin layers” in the soil profile are merged by checking similarity of qc values of adjacent layers. The final soil profile will not contain layers thinner than 15 cm. This avoids the contamination of computations based on the soil profile that would reflect artificial ‘‘thin layers” that cannot even be identified or characterized with the standard cone. The use of the proposed algorithms is demonstrated using a CPT carried out in West Lafayette, Indiana. The ability of the algorithm to logically handle thin layers and separate out state parameters from intrinsic parameters in the SBT classification is highlighted

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

by means of the example. The comprehensive algorithm presented in this paper is robust and can be used in the post-processing of CPT data to give engineers an assessment of the stratigraphic profile at a site. The proposed algorithms can also be used in tandem with soil geospatial variability analyses for a more rational choice of resistance factors in LRFD design. Acknowledgement This material is based upon work supported by the Indiana Department of Transportation through the Join Transport Research Program under project no. SPR 3408 and SPR 4040. The authors are very grateful for this support. References [1] American association of state highway and transport officials. LRFD bridge design specifications. Washington D.C.: AASHTO LRFD Bridge; 2011. p. 1661. [2] Arshad MI, Tehrani FS, Prezzi M, Salgado R. Experimental study of cone penetration in silica sand using digital image correlation. Géotechnique 2014;64(7):551–69 [Thomas Telford Ltd]. [3] Been K, Jefferies MG, Hachey J. The critical state of sands. Géotechnique 1991;41(3):365–81 [Thomas Telford Ltd]. [4] Begemann HKS. The friction jacket cone as an aid in determining the soil profile. In: Proceedings of the 6th international conference on soil mechanics and foundation engineering. Montreal: ICSMFE; 1965. p. 8–15. [5] Bica AVD, Prezzi M, Seo H, Salgado R, Kim D. Instrumentation and axial load testing of displacement piles. Proc ICE – Geotech Eng 2014;167(3):238–52. [6] Cai G, Liu S, Puppala AJ. Comparison of CPT charts for soil classification using PCPT data: example from clay deposits in Jiangsu Province, China. Eng Geol 2011;121(1–2):89–96. [7] Cai G, Liu S, Puppala AJ. Reliability assessment of CPTU-based pile capacity predictions in soft clay deposits. Eng Geol 2012;141–142:84–91. [8] Carraro JAH, Bandini P, Salgado R. Liquefaction resistance of clean and nonplastic silty sands based on cone penetration resistance. J Geotech Geoenviron Eng 2003;129(11):965–76. [9] Carraro JAH, Murthy TG, Loukidis D, Salgado R, Prezzi M. Undrained monotonic response of clean and silty sands. Géotechnique 2007;57(3):273–88. [10] Carraro JAH, Prezzi M, Salgado R. Shear strength and stiffness of sands containing plastic or nonplastic fines. J Geotech Geoenviron Eng 2009;135 (9):1167–78. [11] Chen Q, Wang C, Hsein Juang C. CPT-based evaluation of liquefaction potential accounting for soil spatial variability at multiple scales. J Geotech Geoenviron Eng 2015:4015077 [American Society of Civil Engineers]. [12] Ching J, Wang JS, Juang CH, Ku CS. Cone penetration test (CPT)-based stratigraphic profiling using the wavelet transform modulus maxima method. Can Geotech J 2015;52(12):1993–2007 [NRC Research Press]. [13] Clausen CJ, Aas PM, Karlsud K. Bearing capacity of driven piles in sand, the NGI approach. Front Offshore Geotech 2005:667–81. [14] Dai G, Salgado R, Gong W, Zhang Y. Load tests on full-scale bored pile groups. Can Geotech J 2012;49(11):1293–308. [15] Dijk BFJ Van, Kolk HJ. PT-based design method for axial capacity of offshore piles in clay. In: Frontiers in Offshore Geotechnics II; 2010. p. 555–60, https:// www.crcpress.com/Frontiers-in-Offshore-Geotechnics-II/Gourvenec-White/ p/book/9780415584807. [16] Douglas BJ, Olsen RS. Soil classification using electric cone penetrometer. In: Symposium on cone penetration testing and experience. St. Louis: Geotechnical Engineering Division ASCE; 1981. [17] Eslami A, Fellenius BH. Pile capacity by direct CPT and CPTu methods applied to 102 case histories. Can Geotech J 1997;34(6):886–904. [18] Foye KC, Abou-Jaoude G, Prezzi M, Salgado R. Resistance factors for use in load and resistance factor design of driven pipe piles in sands. J Geotech Geoenviron Eng 2009;135(1):1–13. [19] Foye KC, Salgado R, Scott B. Resistance factors for use in shallow foundation LRFD. J Geotech Geoenviron Eng 2006;132(9):1208–18. [20] Han F, Prezzi M, Salgado R, Zaheer M. Axial resistance of closed-ended steelpipe piles driven in multilayered soil. J Geotech Geoenviron Eng 2016;143 (3):4016102. [21] Jardine R, Chow F, Overy R, Standing J. ICP design methods for driven piles in sands and clays. Thomas Telford Ltd; 2005. [22] Jefferies MG, Davis MP. Soil classification by the cone penetration test: discussion. Can Geotech J 1991;28(1):173–6. [23] Kim D, Salgado R. Load and resistance factors for external stability checks of mechanically stabilized earth walls. J Geotech Geoenviron Eng 2012;138 (3):241–51. [24] Kim D, Salgado R. Load and resistance factors for internal stability checks of mechanically stabilized earth walls. J Geotech Geoenviron Eng 2012;138 (8):910–21. [25] Kim K, Salgado R, Lee J, Paik K. Load tests on pipe pile for development of CPTbased design method; 2002, http://docs.lib.purdue.edu/jtrp/140/.

83

[26] Kolk HJ, Baaijens AE, Senders M. Design criteria for pipe piles in silica sands. In: Frontiers in offshore Geotechnics; 2005. p. 711–6, https://www.crcpress.com/ Frontiers-in-Offshore-Geotechnics-Proceedings-of-the-InternationalSymposium/Gourvenec-Cassidy/p/book/9780415390637. [27] Larsson R, Mulabdic M. Piezocone tests in clay. SGI REPORT; 1991. [28] Lee J, Salgado R. Estimation of bearing capacity of circular footings on sands based on cone penetration test. J Geotech Geoenviron Eng 2005;131 (4):442–52. [29] Lee J, Salgado R, Carraro JAH. Stiffness degradation and shear strength of silty sands. Can Geotech J 2004;41(5):831–43 [NRC Research Press Ottawa, Canada]. [30] Lee J, Salgado R, Paik K. Estimation of load capacity of pipe piles in sand based on cone penetration test results. J Geotech Geoenviron Eng 2003;129 (5):391–403. [31] Lehane BM, Schneider JA, Xu X. The UWA-05 method for prediction of axial capacity of driven piles in sand. In: Frontiers in Offshore Geotechnics; 2005. p. 683–9. [32] Li J, Cassidy MJ, Huang J, Zhang L, Kelly R. Probabilistic identification of soil stratification. Géotechnique 2016;66(1):16–26 [Thomas Telford Ltd]. [33] Olsen R, Mitchell J. CPT stress normalization and prediction of soil classification. In: Proceedings of international symposium on cone penetration testing. Linköping, Sweden: SGI; 1995. [34] Paikowsky SG. Load and resistance factor design (LRFD) for deep foundations. Transport Research Board; 2004. [35] Puppala A, Wattanasanticharoen E, Hoyos LR, Satyanarayana R. Use of cone penetration test (CPT) results for accurate assessments of pile capacities. In: Proceedings of NICE2002: 9th international conference on piling and deep foundations. p. 75–580. [36] Ramsey N. A calibrated model for the interpretation of cone penetration tests (CPTs) in North Sea quaternary soils. In: Offshore site investigation and geotechnics’ diversity and sustainability. Society of Underwater Technology; 2002. [37] Robertson PK. CPT-based soil behaviour type (SBT) classification system – an update. Can Geotech J 2016. cgj-2016-0044. [38] Robertson PK, Campanella R. Liquefaction potential of sands using the CPT. J Geotech Ellipsis 1985;I(3):384–403. [39] Robertson PK. Soil classification using the cone penetration test. Can Geotech J 1990;27(1):151–8. [40] Robertson PK. Estimation of minimum undrained shear strength for flow liquefaction using the CPT. In: Earthquake geotechnical engineering. Balkema, Rotterdam: CRC Press; 1999. p. 1021–8. [41] Robertson PK. Interpretation of cone penetration tests – a unified approach. Can Geotech J 2009;46(11):1337–55. [42] Robertson PK. Soil behaviour type from the CPT : an update. In: Second international symposium on cone penetration testing, Huntington beach, California. [43] Robertson PK, Campanella RG, Gillespie D, Rice A. Seismic CPT to measure in situ shear wave velocity. J Geotech Eng 1986;112(8):791–803. [44] Robertson PK, Wride CE (Fear). Evaluating cyclic liquefaction potential using the cone penetration test. Can Geotech J 1998;35(3):442–59 [NRC Research Press]. [45] Salgado R, Bandini P, Karim A. Shear strength and stiffness of silty sand. J Geotech Geoenviron Eng 2000;126(5):451–62 [American Society of Civil Engineers]. [46] Salgado R, Boulanger RW, Mitchell JK. Lateral stress effects on CPT liquefaction resistance correlations. J Geotech Geoenviron Eng 1997;123(8):726–35 [American Society of Civil Engineers]. [47] Salgado R, Kim D. Reliability analysis of load and resistance factor design of slopes. J Geotech Geoenviron Eng 2014;140(1):57–73 [American Society of Civil Engineers]. [48] Salgado R, Mitchell JK, Jamiolkowski M. Cavity expansion and penetration resistance in sand. J Geotech Geoenviron Eng 1997:344–54 [American Society of Civil Engineers]. [49] Salgado R, Prezzi M. Computation of cavity expansion pressure and penetration resistance in sands. Int J Geomech 2007;7(4):251–65 [American Society of Civil Engineers]. [50] Sanglerat G, Nhim TV, Sejourne M, Andina R. Direct soil classification by static penetrometer with special friction sleeve. In: Proceedings of the first European symposium on penetration testing. p. 5–7. [51] Schmertmann JH. Guidelines for cone penetration test: performance and design. Florida: Gainesville; 1978. [52] Schneider JA, Hotstream JN, Mayne PW, Randolph MF. Comparing CPTU Q – F and Q – D u 2 /r v0 0 soil classification charts. Géotech Lett 2012;2(4): 209–15. [53] Schneider JA, Randolph MF, Mayne PW, Ramsey NR. Analysis of factors influencing soil classification using normalized piezocone tip resistance and pore pressure parameters. J Geotech Geoenviron Eng 2008;134(11):1569–86 [American Society of Civil Engineers]. [54] Scott B, Kim BJ, Salgado R. Assessment of current load factors for use in geotechnical load and resistance factor design. J Geotech Geoenviron Eng 2003;129(4):287–95 [American Society of Civil Engineers]. [55] Senneset K, Sandven R, Janbu N. Evaluation of soil parameters from piezocone tests. Transp Res Rec 1989:1235. [56] Tumay MT. Field calibration of electric cone penetrometers in soft soils. Baton Rouge; 1985.

84

E. Ganju et al. / Computers and Geotechnics 90 (2017) 73–84

[57] Wang Y, Huang K, Cao Z. Probabilistic identification of underground soil stratification using cone penetration tests. Can Geotech J 2013;50(7):766–76 [NRC Research Press]. [58] Xu X, Lehane BM. Pile and penetrometer end bearing resistance in two-layered soil profiles. Géotechnique 2008;58(3):187–97 [Thomas Telford Ltd].

[59] Xu X, Schneider JA, Lehane BM. Cone penetration test (CPT) methods for endbearing assessment of open- and closed-ended driven piles in siliceous sand. Can Geotech J 2008;45(8):1130–41 [NRC Research Press]. [60] Zhang Z, Tumay MT. Statistical to fuzzy approach toward CPT soil classification. J Geotech Geoenviron Eng 1999;125(3):179–86 [American Society of Civil Engineers].