The possibility of a statistical determination of characteristic shear parameters from triaxial tests

Abstract The shear strength of soils, commonly represented in terms of friction angle and cohesion, is an important input parameter for all geotechnical calculations. It can be assessed in the laboratory for example by triaxial tests. From the laboratory test results, the characteristic values for friction angle and cohesion have to be determined – according to the standards – as conservative estimation of the mean value. Usually the shear strength is calculated as linear regression in the s‐t‐representation. This leads to slightly higher values than the evaluation in the more appropriate σ₁‐σ₂‐representation. Friction angle and cohesion are usually assumed to be statistically independent variables, resulting in unrealistically low characteristic shear strengths. Friction angle and cohesion are obviously correlated, since they are parameters of one linear regression. Confidence hyperbolas can be determined by taking into account the statistical dependency of friction angle and cohesion. A linearization of the confidence hyperbolas provides an upper and lower limit of the shear parameters. Typical examples are used to demonstrate the determination of the characteristic shear strength from the laboratory test results with the above mentioned approaches. Moreover we show, that the characteristic shear strength can be determined more precisely if more lateral stress levels (e.g. six) are performed instead of the conventional three levels.