USE OF STATIC CONE PENETRATION TEST FOR CLASSIFICATION OF LOCAL SOILS

Transcription

1 USE OF STATIC CONE PENETRATION TEST FOR CLASSIFICATION OF LOCAL SOILS Abdul Karim M. Zein * ABSTRACT: This paper presents an updated and modified version of a previously developed numerical soil classification method based on analysis of data from the static cone penetration test CPT. Relevant data was obtained in the study for over 900 samples representing various soil types collected from many sites located in different parts of Sudan. Classification of studied soils was made according the Unified System for Classifying Soils (USCS) using the results of standard tests performed in the laborary on representative samples. The cone resistance q c and friction ratio R f were obtained for each soil type from CPT soundings made adjacent boreholes at the same sample depths in each site. The discriminant concept of statistical analysis was employed for proposing a numerical classification method that assigns a given soil one of five main groups selected on soil type basis following the USCS scheme. A computer program that aumatically predicts the type of penetrated soil according the method suggested and gives indication of relative density of sandy or silty soils and consistency of clays was also developed in this study. 1. INTRODUCTION Soil classification is an important aspect in the fields of geotechnical engineering and it simply means the placement of a soil in one of a number of soil groups on the basis of the grading of its coarser particles and the plasticity of finer particles. There is a distinction between the terms soil description and soil classification. A full soil description gives detailed information on the grading, plasticity, color, and particle characteristics, as well as on the fabric, the state of bedding, nature of discontinuities and strength condition in which it occurs in soil sample, borehole or exposure. On the other hand, a soil classification places a soil in a limited number of groups on the basis of grading and plasticity of a disturbed or undisturbed sample. To apply any of the known soil classification systems it is necessary perform certain laborary tests such as the grain size distribution and consistency limits on representative soil samples taken from the various strata. However, in certain cases e.g. during the preliminary stages of site investigation, testing might not be possible due some restraints such as time, lack of suitable apparatus or limitation on available funds. This has lead previous workers consider developing faster and cheaper, but probably less accurate, indirect soil classification methods such as those based on the static cone penetration test (CPT) method. As in many other countries there is a need in Sudan for developing a quick and reliable method for local soils classification especially in urban regions where development projects are normally foreseen or very remote areas where drilling and sampling techniques may not be available. In this paper the author is suggesting a numerical soil classification method based on analysis of considerable amount of CPT data compiled carefully from many soil investigation and research work reports done during the last three decades. As a background for the study a brief review of the previous research works carried out worldwide and locally on the use of CPT for soil classification is presented in the following section. 2. SOIL CLASSIFICTION METHODS BASED ON CPT; A REVIEW It has been reported by various authors e.g. Begemann [1], Schmertmann [2] and Sanglerat [3] that the CPT data may be used for a preliminary soil classification if the tests are performed by a penetrometer fitted with a special sleeve for measurement of local soil friction. Such a penetrometer permits measuring separately the two parameters i.e. the cone resistance q c and the local friction f s by the CPT machine at any depth of the penetrated soil material. Schmertmann [2] stated that either q c or f s alone does not give sufficient information guess the type soil tested and suggested the introduction of the f s /q c ratio termed as the friction ratio R f as a third soil parameter. With this R f and the q c the type of penetrated soil may be identified. Several methods have been proposed and reported in published literature for the classification of soils using the static CPT results. * Associate Prof. BRRI,U. of K., Sudan August 2003 BRRJ Vol. 5 23

2 Begemann [1] and Schmertmann [2] indicated that there is a relationship between the soil type and CPT data and proposed the use of charts in which the cone resistance q c is plotted against the local friction, f s [1] and the friction ratio R f [2] for different types of soil material. From an extensive experimental research made on various soils in France, Sanglerat [3] confirmed the findings of Begemann and Schmertmann and suggested that generally R f is in the order of 5-6 % in clays and % in sands of varying degree of compactness. Douglas and Olsen [4] were probably the first propose a graphical soil classification method based on CPT data measured by an electric cone penetrometer. Their chart was divided in several zones representing different soil groups. Jones and Rust [5] used a piezocone device develop a soil classification chart in which the excess pore water pressure is plotted against the net cone resistance. The latter parameter was obtained by subtraction of the tal overburden pressure from the tal recorded cone resistance. This method has the advantage of giving additional information on the relative density in coarse grained soils and the consistency in fine grained soils. Robertson et al [6] developed a soil classification chart based on CPT data measured by a piezocone device with the cone resistance values corrected for development of pore water pressure. The chart identifies different soil types divided in 12 separate zones. Few years later, Robertson [7] proposed a refinement of the previous method by plotting a normalized cone resistance against a normalized friction ratio obtained from formulae established for correcting the effects of tal and effective overburden pressures. Fellenius and Eslami [8] developed a classification method in which soils were grouped in five different types. The effective cone resistance was obtained by subtracting pore water pressure and plotted against the friction resistance of the soil. These CPT classification methods may prove be quite useful when applied in some soils different from those for which they have been developed but differences may well be indicated in other locations because of their empirical nature. It is therefore recommended examine the validity of any system before being used in countries where the experience on the interpretation of the CPT data is not adequate. This may be achieved by making comparison of soil classification made according the CPT methods against those revealed from laborary testing of the same soil type. In Sudan, the first CPT machine was imported in 1977 and used by a Dutch Consulting firm in the site investigation for Jonglie Canal project in southern Sudan. That investigation was carried out by BRRI (Building and Road Research Institute) of University of Kharum and the Dutch engineers during 1977 and The work on that project offered a good opportunity for BRRI gain experience on the use and interpretation of the CPT data and provided considerable data for carrying out comparison studies between the CPT and the standard tests for some local soil. A soil classification method was developed since 1980 from analysis of CPT and standard laborary test results for various soil samples taken from southern Sudan and other sites in the Greater Kharum area. A detailed description of the CPT soil classification method for local soils is given elsewhere [9] but a brief account on the same is outlined here. Zein and Ismail [10] analysed CPT data points pertaining soil types that had been tested in the laborary determine their grain size distribution and consistency characteristics. All soil samples tested were classified from test results and divided in four main soil groups namely clays, silty and sandy clays, clayey sands and silt-sand mixtures, and sands. Static cone penetration tests were carried out using a 10 ns capacity sounding machine fitted with a mechanical adhesion jacket cone (Begemann s tip) and the q c and R f values were obtained at the depths of the soil samples considered in the analysis. It was noted from plotting a combined q c -soil type- R f chart that each soil group tends occupy a certain region in the plot, though overlap between the groups can however be observed. To enable classification of a soil sample for which only the CPT data is available, the zone occupied by each soil group should be defined. A statistical approach of data analysis known as the descriminant method [11] was chosen differentiate mathematically between the four zones of the CPT data corresponding the main soil groups considered in the study. This method of statistical analysis is briefly explained in the following section. 3. MODIFIED CPT DESCRIMINANT SOIL CLASSIFICATION METHOD August 2003 BRRJ Vol. 5 24

3 3.1 Aims of Study The method described in Section 3 and published as a BRRI current paper in 1982, was proposed for application by interested researchers and engineers for the classification of local soils using the CPT data. Since then, a considerable data was made available from laborary and CPT results of several site investigation projects and research studies carried out at BRRI. The availability of such data encouraged the author consider updating and modifying the originally developed descriminant soil classification method meet the current requirements of practicing engineers and research workers by satisfying the following objectives: To improve the degree of classification accuracy by including in the analysis the soil test data obtained in the last two decades (i.e. from ). To consider division of CPT data in five soil groups instead of four in previous method so as be more specific in soil classification of the various investigated sites. To develop computer software that simplifies and speeds up the analytical procedure and computations involved in the updated soil classification method. To include in the new version of soil classification method some useful information on the relative density of cohesion-less soils and the consistency of cohesive soils. 3.2 Analysis of Soil Classification Test Results and CPT Data A great effort was made for collection of relevant data that include soil classification test results for samples representing various soil types from many sites distributed in widely different parts of the country. Atterberg limits and/or grain size distribution tests were performed in the laborary on more than 900 representative soil samples taken from different borehole depths in these sites. Only the boreholes drilled adjacent cone penetration test points were included in this study enable sound comparison between CPT data and laborary classification test results. Classification of all soil samples tested was made in accordance with the Unified System for Classifying Soils (USCS). Cone penetration tests 11 = Variance of the variable x 1 (q c ) = var. of q c / sample size 4 = Variance of the variable x 2 (R f ) 22 were made by two mechanical machines with 100 and 200 kn capacities equipped with standard adhesion or friction jacket cone which facilitates separate measurements of cone resistance q c and local friction f s of penetrated soil in 200mm depth intervals. The soil types covered in the present analysis were divided in the following five main groups using the same terminology of the USCS for the purpose of statistical analysis and subsequent classification on the basis of CPT only: i) Clays of high plasticity (CH) ii) Clays of low plasticity (CL) iii)silty soils of low high compressibility (ML, MH) vi) Clayey and silty sands (SC and SM) and v) Poorly and well graded clean sands (SP, SW). The statistical discriminant method followed for analyzing the CPT data and developing the soil classification scheme is briefly described here. In this analysis, the term soil population has the same meaning of the term soil group, and both describe one set of data having similar characteristics. In all soil populations an assumption was made that the distribution of CPT data is normal and that all populations have identical covariance matrices. The descriminant function, denoted by X l, for each population is given as follows: X l = j k lj x k - ½ j k lj l k (1) l = number of population = 1, 2, 3, jk = jk th element of the inverse of the common covariance matrix of all populations, j = mean of the measured variable values, and x k = corresponds the common variables in each population which are in our case the cone resistance q c and the friction ratio R f of the tested soil. The covariance matrix is be computed for each population using its elements defined according the following relationships: = (2) = (var. of R f ) / sample size 4 12 = 21 = Covariance of q c and R f = covar. q c & R f / sample size 4 August 2003 BRRJ Vol. 5 25

4 The measured CPT data pertaining the five different soil populations adopted in this study may be used for deriving the descriminant functions. Each population has a certain function of which parameters are be estimated from the sample data that is known for certain come from that population. The variance and covariance values for the q c and R f variables were found from the data corresponding each soil group and the results obtained are summarized in Table 1 which also gives the sample sizes considered and the mean q c and R f values. Table (1) Summary of CPT data used in descrimenant analysis for soil classification From the data listed in this Table 1, the pooled or common covariance matrix was calculated for the five populations using equation 2 and the numerical values of its elements were found be as follows: Soil group Statistical data Clays of high plasticity (CH) Clays of low plasticity (CL) Silty soils (ML or MH) Clayey or silty sands (SC or SM) (3) Clean Sands (SP or SW) Data size Mean ( 11 ) (MN/m 2 ) q c Variance of q c Mean R f ( 12) (%) Variance of R f Covariance of q c and R f The inverse of the pooled covariance matrix is the one be used in the computations of the descriminant functions of the soil groups according equation 1 and the numerical values of its elements are given in the following matrix: = = = 21 (4) The descriminant function of each soil group shall then be computed according equation 1 using the mean values 11 and 12 of the two variables x 1 (q c ) and x 2 (R f ) respectively and the numerical values of the inverse of the pooled covariance matrix given by equation 4. Substitution of the appropriate values from Table 1 in equation 1 for the five different soil groups reveals the following descriminant functions: X CH = 0.35q c R f 8.31 (5.1) X CL = 0.39q c R f 5.39 (5.2) X ML, MH = 0.41q c R f 4.86 (5.3) X SC, SM = 0.58q c R f 5.87 (5.4) X SP, SW = 0.70q c R f 5.99 (5.5) The units of q c and R f in equations are MN/m 2 and % respectively. The use of the descriminant functions in equations for the CPT based classification of soils is exactly the same as described in the previous method and is briefly outlined here. A soil sample of known q c and R f values but of uncertain type is allocated the nearest population where nearness here is a measure of probability. The nearest population is that from which a greater likelihood of the sample is coming. Therefore, the sample should be allocated whichever population gives the greatest value X l in equation 1. Supposing that for an arbitrary point at any depth, only q c and R f values are known and it is required predict which type of soil could be encountered at that depth according the descriminant soil classification method. Substituting the values of q c and R f in equations gives five numerical values known as the descriminants and the sample in question is simply allocated the soil group from which the greatest numerical value is obtained. As an example, three soils types with different q c and R f values as assumed in Table 2 would give different numerical values by substituting the appropriate q c and R f values in equations 5.1 through 5.5. The highest numerical values of the discriminants (printed in bold) for the three samples indicate the August 2003 BRRJ Vol. 5 26

5 correct classification of each soil according the new proposed method as shown in the last column of Table 2. Soi l No Table (2) Examples of using the modified CPT soil classification method qc MN/m 2 R f (%) Numerical values of disciminants obtained by substitution in equations CH CL 5.3 ML/ MH 5.4 SC/ SM 5.5 SP/ SW Correct soil classifica tion based on CPT data CH MlorMH SPorSW To facilitate a continuous profiling of soil strata at any CPT point, a short computer program was made enable computations of discriminant values according equations 5.1 through 5.5 for the depth at which the q c and R f values are normally measured. The program listed in Appendix A enables instant prediction of the soil type for CPT soundings at any depth of soil formation. A useful feature was added the proposed soil classification method provide indicative information on the degrees of consistency in cohesive soils and relative density in cohesionless soils. The correlation relationships developed by Terzaghi and Peck [12] between the standard penetration test N-value on one hand and relative density of sandy soils and consistency of clay soils on the other were used as reference in this study. The N value may be replaced by the equivalent cone resistance q c by using some available correlation between the two parameters established by various authors for different soil types. In a recent study (13) an empirical relationship was developed between the SPT N value (blows/305mm), the cone resistance q c (kg/cm 2 ) and the friction ratio R f (%) for locally occurring soils as follows: q c = (N/ R f ) (N/R f )+10.3 (6) The values of q c which correspond N values were estimated according equation 6 using the average R f values from Table 1 and are given in Table 3 for the five main soil groups adopted in this study. Table (3) Consistency of clays and relative density of sands and silts based on CPT Clay soils Consistenc y N valu e Very soft <2 <1. 3 Soft Medium 4 8 Stiff 8 15 Very stiff Equivalent q c values in MN/m 2 CH CL Relative density Hard >30 >4. 7 < >6. 0 Sandy and silty soils Very Loose Loose Medium Dense Very dense N Valu e Equivalent q c values in MN/m 2 ML/ MH <4 < >50 >9. 4 SC/ SM SP/ SW <1.9 < >10. 5 Therefore, the CPT soil classification method may be used not predict the soil type only but moreover evaluate roughly some of its physical and engineering properties such as density and shear strength. The computer program outlined in Appendix A was also designed such that it provides information on this useful and complimentary feature of soil classification. 4. CONCLUSIONS An analytical empirical method based on the CPT data was presented in this paper and proposed for preliminary classification of local soil types. The method was developed by comparison made between soil classification based on the USCS and the CPT cone resistance q c and friction ratio R f for over 900 soil samples representing widely different soil types and conditions in Sudan. Analysis of the CPT data was made by the statistical descriminant method derive mathematical functions (equations ) representing five main groups adopted for the purpose of soil classification. According the proposed classification method a soil of known q c and R f but of uncertain type is allocated the nearest group from which a greater likelihood of the sample is coming i.e. that gives the greatest value X l in equations In addition, the modified method gives a rough indication of the consistency of clays and relative density of sandy and silty soils based on correlation between their degrees and measured q c values. A computer program was prepared and listed in Appendix >14. 8 August 2003 BRRJ Vol. 5 27

6 A enable classification of penetrated soils according this method based on the CPT data. 5. ACKNOWLEDGEMENT The author acknowledges with gratitude the assistance offered him by Engineer Asher Rifaat, M.Sc. student at BRRI, for collection some of the data used for analysis in this study and Engineer Hisham Osman, M.Sc. student at BRRI, for the preparation of computer program listed in Appendix A. 6. REFERENCES [1] Begemann H.K.S.. The Dutch static penetration test with the adhesion jacket cone. Lab. Groundmech., Delft, Netherlands, 13(10): 1-86, (1969). [2] Schmertmann, J.H.. Guidelines for CPT performance and design. Report prepared for Fedral Highway Administration, Washingn D.C., (1977). [3] Sanglerat G.G.J.. The penetrometer and soil exploration. Development in Geotechnical Engineering. Elsevier Publishing Co., Amsterdam, (1972). [4] Douglas B.J. and Olsen, R.S., Soil classification using electric cone penetrometer. Symp. on cone penetration testing and experience. Geotechnical Engineering Division of ASCE, St. Louis, pp: , (1981). [5] Jones, G. A. and Rust, A. Piezometer penetration testing, CPUT. Proc. 2 nd European Symp. on Penetration Testing, ESPOT-2 Amsterdam, Vol. 2, pp , (1982). [6] Robertson, P.K., Campenalla, R.G. and Wightman, A., SPT-CPT correlations. J. Geotechnical. Eng. Division, ASCE, Vol. 109, GT11, pp: , (1983). [7] Robertson, P.K., Soil classification using the cone penetration test. Canadian Geotechnical Journal, Vol. 3, No. 1, pp , (1990). [8] Fellenius, B. H. and Eslami, A., Soil Profile interpreted from CPTu data. Proc. Year 2000 Geotechnics Geotchnical Engineering Conference, Asian Institute of Technology, Bangkok, 18p. (2000). [9] Zein, A. K. M., Correlation between static cone penetration and recognized standard test results for some local soils. M.Sc. Thesis, Civil Eng. Dept., U. of K., (1980). [10] Zein, A.K.M., and Ismail, H.A.E.,Use of static cone penetration test for soil classification. BRRI Current Paper Publication CP1/81, (1981). [11] Kendal, M., Multivariate analysis. Griffin and Charles Publishing Co. (1975). [12] Terzaghi, K. and Peck R.B., Soil mechanics in engineering practice, J.Wiley and Sons Inc., NY., (1948). [13] Zein, A. K. M., Development and evaluation of some empirical methods of correlation between CPT and SPT. Building and Road Research Journal, Vol. 4, 16-28, (2002). Appendix A: Computer program for soil classification from CPT data. REAL RF,QC,D,X1,X2,X3,X4,X5 INTEGER N,C C=0 PRINT *,"INTER THE NUMBER OF VALUE" READ*,N 10 PRINT *,"INTER THE VALUE OF DEPTH" READ*,D PRINT *,"INTER THE VALUE OF qc" READ*,QC PRINT *,"INTER THE VALUE OF Rf" READ*,RF X1=0.35*QC+2.47*RF-8.31 X2=0.39*QC+1.87*RF-5.39 X3=0.41*QC+1.73*RF-4.86 X4=0.58*QC+1.59*RF-5.87 X5=0.70*QC+1.12*RF-5.99 IF((X1.GT.X2).AND.(X1.GT.X3).AND.(X1.GT. X4).AND.(X1.GT.X5)) GOTO20 IF((X2.GT.X1).AND.(X2.GT.X3).AND.(X2.GT. X4).AND.(X2.GT.X5)) GOTO30 IF((X3.GT.X1).AND.(X3.GT.X2).AND.(X3.GT. X4).AND.(X3.GT.X5))GOTO40 IF((X4.GT.X1).AND.(X4.GT.X3).AND.(X4.GT. X3).AND.(X4.GT.X5))GOTO50 IF((X5.GT.X1).AND.(X5.GT.X2).AND.(GT.X3). AND.(X5.GT.X4))GOTO60 20 IF(QC.LT.1.3) THEN PRINT*,"CH -Very soft at",d IF(QC.GT.1.3.AND.QC.LT.1.6) THEN August 2003 BRRJ Vol. 5 28

7 PRINT*,"CH - soft at",d IF(QC.GT.1.6.AND.QC.LT.2.1)THEN PRINT*,"CH - Medium at",d IF((QC.GT.2.1).AND.(QC.LT.2.9))THEN PRINT*,"CH - Stiff at",d IF(QC.GT.2.9.AND.QC.LT.4.7)THEN PRINT*,"CH - Very stiff at",d PRINT*,"CH - Hard at",d 30 IF(QC.LT.1.4) THEN PRINT*,"CL -Very soft at",d END IF IF(QC.GT.1.4.AND.QC.LT.41.7) THEN PRINT*,"CL - soft at",d IF(QC.GT.1.7.AND.QC.LT.2.4) THEN PRINT*,"CL - Medium at",d IF(QC.GT.2.4.AND.QC.LT.3.6)THEN PRINT*,"CL - Stiff at",d IF(QC.GT.3.6.AND.QC.LT.6.0)THEN PRINT*,"CL - Very stiff at",d PRINT*,"CL - Hard at",d END IF 40 IF(QC.LT.1.8)THEN PRINT*,"ML/MH -Very Loose at",d IF(QC.GT.1.8.AND.QC.LT.2.9)THEN 50 IF(QC.LT.1.9) THEN PRINT*,"SC/SM -Very Loose at",d IF(QC.GT.1.9.AND.QC.LT.3.2)THEN PRINT*,"SC/SM - Loose at",d IF(QC.GT.3.2.AND.QC.LT.7.2)THEN PRINT*, "SC/SM-Medium at", D IF(QC.GT.7.2.AND.QC.LT.10.5)THEN PRINT*,"SC/SM - Dense at",d PRINT*,"SC/SM - Very dense at",d 60 IF(QC.LT.2.5) THEN PRINT*,"SP/SW -Very Loose at",d IF(QC.GT.2.5.AND.QC.LT.4.6)THEN PRINT*,"SP/SW - Loose at",d IF(QC.GT.4.6.AND.QC.LT.10.6)THEN PRINT*,"SP/SW - Medium at",d IF(QC.GT.10.6.AND.QC.LT.14.8)THEN PRINT*,"SP/SW - Dense at",d PRINT*,"SP/SW - Very dense at",d 100 C=C+1 IF(C.LT.N)GOTO10 STOP END PRINT*,"ML/MH - Loose at",d IF(QC.GT.2.9.AND.QC.LT.4.6)THEN PRINT*,"ML/MH - Medium at",d IF(QC.GT.6.4.AND.QC.LT.9.4)THEN PRINT*,"ML/MH - Dense at",d PRINT*,"ML/MH - Very dense at",d August 2003 BRRJ Vol. 5 29