Laboratory evaluation of permeability coefficient relationships for sandy soils

Assistant Professor¡ Soil Conservation and Watershed Management Institute (SCWMRI)¡ Tehran¡ Iran. The saturated hydraulic conductivity is one of the most important and widely used geotechnical parameters. This parameter depends on properties of the fluid¡ pore size distribution¡ and characteristics of the solid surfaces. Because the latter two are not necessarily constant¡ the hydraulic conductivity may vary significantly. So far several empirical relationships for estimation of the hydraulic permeability with respect to the type of materials have been presented. In this research seven applicable method including Terzaghi¡ Kozeny-Carman¡ Chapuis and Aubertin¡ modified Navfac¡ Shahabi et al.¡ Mbonimpa et al. and Chapuis were evaluated. Four samples of sandy soils (R1 to R4) with a rather wide range of particle sizes were prepared and were compacted within the Plexiglas box. After providing the head of 0.5¡1 and 1.5 meters; flow discharge and static pressure were measured which result phreatic and isopotential lines extraction. By applying the results of the model¡ the accuracy of some common methods used in the estimation of saturated hydraulic conductivity of sandy soils was evaluated. For the samples (R1 to R4)¡ the coefficient of permeability was 0.0051¡ 0.048¡ 0.076¡ and 0.19 cm/s respectively. Due to the limited range of input parameters¡ the method of Shahabi et al. can be used only for the R4 sample; so this method was discarded from statistical analysis. Comparing the results of laboratory test with common methods shows the methods of KozenyCarman and Chapuis and Aubertin were more accurate according to the highest R2 and lowest RE and RMSE. These methods were based on effective parameters and the results of accurate laboratory studies. The method of Mbonimpa et al. (2002) indicates the maximum deviation from observed values which may be due to the sensitivity of this model to void ratio (e). Also the results shows by increasing of void ratio the error of Mbonimpa et al. method increases.