APPROPRIATE PARAMETERS FOR PREDICTION OF SWELLING PRESSURE OF EXPANSIVE CLAYS

Appropriate IGC 009, Guntur, Parameters INDIA for Prediction of Swelling Pressure of Expansive Clays APPROPRIATE PARAMETERS FOR PREDICTION OF SWELLING PRESSURE OF EXPANSIVE CLAYS Sudha Rani Associate Professor, Dept. of Civil Engineering, S.V. University, Tirupati 517 50, Andhra Pradesh, India. E-mail: sudhajawahar@gmail.com K. Mallikarjuna Rao Professor, Dept. of Civil Engineering, S.V. University, Tirupati 517 50, Andhra Pradesh, India. E-mail: kmr_svuce@yahoo.com ABSTRACT: Predetermination of swelling characteristics (Swelling Pressure, Swell Potential and Swell Index) is essential for safe and cost-effective design of structures resting on expansive soils. In this paper an attempt has been made to develop a correlation for Swelling Pressure accounting both soil state and soil type representative parameters. The soil state is reflected by environmental factors namely Moisture Content, Dry Density and Surcharge Pressure whereas the soil type is reflected by the compositional parameters namely Liquid Limit and Plasticity Index. Four series of free Swell Odeometer tests were conducted on four soil samples under different soil state conditions (varying initial Moisture Content, initial Surcharge Pressures and initial Dry Density). A regression model was developed relating logarithm of Swelling Pressure with three of the soil state parameters and two of the soil type parameters. Relative influence of each of the five parameters on Swelling Pressure is estimated statistically by partial correlation coefficients and its performance is verified with seventy soils data from literature in the form of graphs. The study revealed that all the five parameters have significant influence on Swelling Pressure. 1. INTRODUCTION Substantial literature has unequivocally documented and spelled out the severity and extent of damages inflicted by soil deposits of swelling nature in many parts throughout the world (Jones & Jones 1987, Chen 1975, Dhowian et al. 1990, Abdul Jawad et al. 199). Although it has been many years since the swelling phenomenon has been fully recognized, still there is no established method of measuring swelling characteristics of clays. Attempts have been made by several investigators to correlate Swelling Pressure with various index properties (Komarnik & David 1969, Vijayvergiya & Ghazzaly 1973, Nayak & Christensen 1974, Chen 1975, Brackley 1975, Weston 1980, Mowafy et al. 1985, Mallikarjuna Rao 1988 & Yusuf Erzin & Orhan Erol 004). Careful study of the proposed equations reveal that the investigators used either Liquid Limit, Clay Fraction, Activity or one of the placement conditions namely natural Moisture Content, initial Void Ratio and initial Dry Density for development of correlations. The factors considered for development of correlations may be grouped into two broad categories viz., compositional and environmental factors. Liquid Limit, Plasticity Index falls under compositional factors which are known to reflect compositional factors like type and amount of clay minerals, activity and pore water chemistry. whereas initial Void Ratio, initial Moisture Content, initial Dry Density and initial Surcharge Pressure falls under environmental factors. In this investigation an attempt has been made to develop correlations for Swelling Pressure accounting for all the above-identified factors by conducting a series of laboratory tests on remoulded soil samples.. METHODS AND MATERIALS Expansive clayey soils were collected from four different places namely Panidham, Kurnool, Guntur and Madanapalli of Andhra Pradesh and are designated as SS1, SS, SS3 and SS4 for further reference. The required amount of soil is collected from the trial pits at a depth of 1.5 m below the ground level since the topsoil is likely to contain organic matter and other foreign materials. The soils are collected carefully so that the soil samples are fairly homogeneous. The soil is air dried after transporting the same to the laboratory and is pulverized with a wooden mallet. The soil so prepared is sieved through a 4.75 mm sieve and stored in storage bins in the laboratory for further testing. The index properties of these soils are presented in Table 1. All the soils do contain fine fraction of about 95% and are clay of intermediate to high plasticity. 3. TESTS AND RESULTS Four series of Free Swell Odeometer tests were conducted on all the four soils samples SS1, SS, SS3 and SS4 at different placement conditions accounting a total of 46 test samples in order to study the effect of placement conditions on the Swelling Pressure. The typical test results for soil SS1at different placement conditions are given in Table. The test results obtained are analyzed to study the influence of each of the environmental factors and compositional factors 9

on Swelling Pressure. The variation of Swelling Pressure with each of the placement factors while keeping other two placement factors constant among the three (w L, I P,? d ), and the variation of P S with each of the compositional factors (w L, I P ) at different placement conditions was examined graphically. Typical plots showing the variation of P S with one of the placement factors (? d) and one of the compositional factors (w L ) at different placement conditions are given in Figures 1 &. From the graphical plots showing variation of P S versus placement factors and P S versus compositional factors. it can be concluded that the Swelling Pressure is dependent on all placement conditions namely initial Moisture Content, initial Dry Density and initial Surcharge Pressure and also the Liquid Limit and Plasticity Index of the soil. Placement conditions reflect environmental factors whereas Liquid Limit and Plasticity Index reflect compositional factors. Hence Swelling Pressure can be said to be dependent on both environmental factors and compositional factors and may be expressed as given below. P S = f ((w L, I P,? d, m c, S i ) (1) The Swelling Pressure is observed to bear a nonlinear relationship with all the influencing factors (? d, m c and S i ). The Swelling Pressure is observed to increase more sharply with Dry Density than with m c and S i. Hence no linear relationship between the independent and dependent variable is possible. In order to make the relationship linear, logarithm of Swelling Pressure (Log P S ) and logarithm of initial Dry Density (Log? d ) are considered in the development of relationships. Hence relationship between Swelling Pressure and initial Moisture Content, initial dry density, initial surcharge, Liquid Limit and Plasticity Index can be expressed as given below. Table 1: Properties of Tested Soils Properties SS1 SS SS3 SS4 Gravel 0 0.6 0.4 1 Sand 5 5. 4.4 4.8 Silt + Clay 95 94. 95. 94. Liquid Limit (W L 10 69 56 48 Plastic Limit (W P ) 3 36.9 30.81 1.4 Plasticity Index (I P ) 88 3.1 5.19 3.61 Free Swell Index (FSI) 75 10 90 75 Shrinkage Limit (W S ) 8.5 9 11 13.5 I.S Classification CH CH CH CI Specific Gravity (G s ).75.78.81.87 Degree of Expansion VH H M M *VeryHigh (VH), High (H), Medium (M), Low (L) S. No. Soil Table : Typical Test Results for Swelling Pressure for Soil SS1 Atterberg limits Placement conditions w P w L I P? d (kn/m 3 ) m c S i (kpa) Ps (kpa) 1. SS1 3 10 88 13.50 0 5 330. SS1 3 10 88 14.50 0 5 470 3. SS1 3 10 88 15.50 0 5 680 4. SS1 3 10 88 16.50 0 5 950 5. SS1 3 10 88 16.00 0 5 810 6. SS1 3 10 88 17.00 0 5 110 7. SS1 3 10 88 16.00 16 5 410 8. SS1 3 10 88 16.00 0 5 350 9. SS1 3 10 88 16.00 4 5 3000 10. SS1 3 10 88 16.00 8 5 50 11. SS1 3 10 88 16.00 0 0 710 1. SS1 3 10 88 16.00 0 40 600 13. SS1 3 10 88 16.00 0 60 510 14. SS1 3 10 88 16.00 0 80 430 93

Fig. 1: Swelling Pressure Variation with Initial Dry Density Fig. : Swelling Pressure vs Liquid Limit at Varying Moisture Contents (γ d = 16 kn/m 3, S i = 5 kpa) Log(P S ) = ( a0 + a1( WL) + a( IP) + a3( Log dγ ) + a4( mc) a+ 5( Si) ) The values of a 0, a 1, a, a 3, a 4 and a 5 can be obtained from multiple regression analysis. Microsoft-Excel software provides a subroutine for multiple regression analysis and the same is used here to obtain the regression coefficients a 0, a 1, a, a 3, a 4 and a 5 as well as the regression model and correlation coefficient, R. The regression model so obtained is given below along with R value. Log (P S ) = (( 4.3341) + (0.0071 W L ) + (0.0006 I P ) + (5l.80 Log (? d )) (1.7900 m c ) 0.0037 S i )) (3) Correlation coefficient R = 0.979 Swelling Pressure can be predicted using above equation knowing the placement conditions, Liquid Limit and Plasticity Index. The regression analysis yielded a correlation coefficient of 0.979 indicating good correlation between the variables and the Swelling Pressure. Any attempt to correlate Swelling Pressure with either compositional factors alone or environmental factors alone or any other combination other than the () one presented in equation 1 did not yield any fruitful regression models. Hence the same were not presented here. In order to assess the influence of each of the compositional and environmental factors on prediction of Swelling Pressure partial correlation coefficients are determined. The influence of any given parameter is found out by keeping aside that particular parameter and finding the multiple correlation coefficient thereby the partial correlation coefficient. The partial correlation coefficient is obtained using the following equation r1 i = (1 ((1 R 1 )/(1 R1 i ) )) (4) Where r 1 i = partial correlation coefficient R = Multiple correlation coefficient between x 1 1 (i.e. P S ) and all the independent variables (w L, m c, Log? d, I P and S i ). R1 i = Multiple correlation coefficient between x 1 (i.e. P S ) and all the independent variables except the chosen x i (choosing one among w L, m c, Log? d, I P and S i as influencing parameter and the remaining four variables as independent variables). Table 3 summarizes the regression models developed for logarithm of Swelling Pressure by dropping only one parameter namely Liquid Limit or Initial Moisture Content or Logarithm of Dry Density or Plasticity Index or Initial Surcharge at a time along with multiple correlation coefficients. The regression models are designated as E1 to E6 in that order. The table also includes regression model presented in equation 3 designated as E1, which considers all of the compositional and environmental factors for the purpose of comparison. The partial correlation coefficients are estimated using the equation 4 choosing w L, I P, Log? d, m c and S i as influencing parameters in that order for model. The partial correlation coefficients are 0.4, 0.045, 0.888, 0.943 and 0.870, respectively. Since the partial correlation coefficients are significant for all the three placement factors, it may be concluded environmental factors appear to have a dominating influence on Swelling Pressure than compositional factors. The partial correlation coefficient for Plasticity Index is very low indicating that its influence on Swelling Pressure is not so significant in comparison to other factors. The same is reflected by observing the standard deviation of residuals, where the deviations are more in case of environmental factors when any one of the placement factors is neglected. Hence, it may be concluded that the regression model E1 and E in Table 3 can be expected to have more general applicability. This owes to the fact that these models include all the influencing parameters namely w L, m c, Log? d and S i as evident from partial correlation coefficients. 94

S. No. Model No. Table 3: Regression Models Developed for Prediction of Logarithm of Swelling Pressure (Log P S ) Parameters used Influencing parameter Multiple correlation coefficient Regression equation 1. E1 w L, I P, Log? d, m c, S i 0.979 (( 4.334) + (0.007 w L ) + (0.0006 I P ) + (5.80 (log(? d ))) (1.790 m c ) (0.004 S i )). E I P, Log? d, m c, S i w L 0.965 (( 4.61) + (0.008 I P ) + (5.390 Log(? d )) (1.773 m c ) (0.004 S i )) 3. E3 w L, Log? d, m c, S i I P 0.978 (( 4.336) + (0.008 LL) + (5.69 Log(? d )) (1.79 Mc) (.004 Si)) 4. E4 w L, I P, m c, S i Log? d 0.81 ((1.975) + (0.009 w L ) (0.00 I P ) (1.718 m c ) (0.004 S i )) 5. E5 w L, I P, Log? d, S i m c 0.633 (( 3.949) + (0.005 w L ) + (0.007 I P ) + (4.841 Log (? d ))) (0.00 S i )) 6. E6 w L, I P, Log? d, m c S i 0.837 (( 4.111) + (0.004 w L ) + (0.004 I P ) + (5.069 Log (? d )) (1.384 m c )) 4. VERIFICATION WITH THE REPORTED DATA The applicability of the proposed correlation for Swelling Pressure is assessed by comparing the predicted values of Swelling Pressure for the results reported in this investigation as well as using the test results reported for 76 soils from literature by Komarnik & David (1969), Ramudu (1994), Mallikarjuna Rao (1988) in the literature. The Swelling Pressure so predicted is plotted against observed Swelling Pressure for the results of this investigation and for the reported data. These plots are shown in Figures 3 & 4. The solid line in these plots indicates the line of equality. The points are found to fall close to the line of equality in case of results of this investigation indicating good prediction. This is expected because it is the data used for development of proposed regression model However, for other s data, though many points are falling close to the line of equality, some of the points are dispersed away from line of equality. In other words, prediction is good for many soils but not for all soils. This may be attributed to the fact that coarse fraction which can influence swelling characteristics has not been accounted in the proposed regression model. Fig. 3: Observed vs Predicted Swelling Pressure (Results of Present Investigation) Fig. 4: Observed vs Predicted Swelling Pressure (Others data) 5. CONCLUSIONS The factors considered for development of correlations for prediction of Swelling Pressure PS are grouped into two broad categories viz., compositional and environmental factors. wl, IP falls under compositional factors whereas mc,? d and falls under environmental factors. Regression models were developed relating Log(PS) with w L, I P, Log? d, m c and S i separately as well as in various combinations of these five parameters. w L, Log? d, m c and S i are identified as the parameters having significant influence on Log(P S ) based on the partial correlation coefficient. Regression model E1 and E are found to have more general applicability as both the models include all the influencing parameters namely w L, Log? d, m c and S i as evident from partial correlation coefficients. The Performance of the regression model E1 accounting all the five parameters is found satisfactory when verified by comparing the observed PS and predicted P S values for 70 soils test data reported in literature but not used in the development of the models. 95

REFERENCES Abdul Jawad, S.N., Hammed, R.A., Al-Sulaimani, G.I, Basunbul, I.A. and Safar, M.M. (199). Expansive Soils in Eastern Province of Saudi-Arabia, In: Proc. 7th Int. Conf. on Expansive Soils, Dallas, USA, Vol. 1, 46 431. Brackley, I.J.A. (1975). Swell Under Load, Proc. 6th Reg. Conf. for Africa on SM and FE, Curban, S.S., 65 70. Chen, F.H. (1975). Foundations on Expansive Soils, Elsevier Scientific Publishing Co., Amsterdam. Dhowian, A.W. (1990). Simplified Heave Prediction Model for Expansion Shale, Geotechnical Testing Journal, Vol. 13, 33 333. IS: 1498 1970, Classification and Identification of Soils for General Engineering Purposes, BIS, New Delhi. IS: 70 1977 (Reaffirmed 1997) Part XLI, Measurement of Swelling Pressure of Soils (BIS), New Delhi. Jones, D. Earl and Jones, Karena (1987). Treating Expansive Soils, ASCE, Civil Engineering, Vol. 57, No. 8. Komarnik, A. and David, D. (1969). Prediction of Swelling Pressure of Clays, ASCE, J of SM&FE Div., Vol. 95, SM1, 09 5. Mallikarjuna, Rao (1988). A Proper Parameter for Correlation of Swell Potential and Swell Index for Remould Expansive Clays, M.Tech. Thesis Submitted to Dept. of Civil Engg., J.N.T. University College of Engg., Kakinada, India. Mowafy, M.Y. and Bauer, G.E. (1985). Prediction of Swelling Pressure and Factors Affecting the Swell Behaviour of an Expansive Soils, Transportation Research Record, Vol. 103, pp. 3 8. Nayak, N.V. and Christensen, R.W. (1974). Swelling Characteristics of Compactedexpansive Soils, Clays and Clay Minerals, Vol. 19, No. 4, pp. 51 61. Ramudu, M. (1994). Prediction of Swelling Characteristics of Remoulded Expansive Soils, M.Tech Thesis Submitted to S.V.U. College of Engineering Tirupati. Ranganatham, B.V. and Satyanarayana, B. (1969). A Rational Method of Predicting Swelling Potential, for Compacted Expansive Clays, Proc. 6th Int. Conf. SM and FE, Canada, Vol. 1, pp. 9 96. Vijayvergiya, V.N. and Ghazzaly, O.I. (1973). Prediction of Swelling Potential for Natural Clays, Proc. 3rd Int. Conf. on Expansive Soils, Haifa, Israel, Vol. 1, pp. 7 36. Weston, D.J. (1980). Expansive Roads Treatment for Southern Africa, Proc. Fourth. Int. Conf. on Expansive Soils, Denver, Vol. 1, 339 360. Yusuf, Erzin and Orhan, Erol (004). Correlations for Quick Prediction of Swelling Pressure, Electronic Journal of Geotechnical Engineering, Vol. 9F, Paper No. 0476. 96