Determination of in-situ engineering properties of soil using response surface methodology

Since in-situ soils seldom behave like remolded laboratory soils or disturbed field samples, it is important to "identify" or "calibrate" the engineering properties of field soil by means of in-situ tests. A response surface methodology based on an orthogonal regression in the parameter space has been developed to "identify" engineering properties of any material based on in-situ tests. The proposed methodology worked very well (i.e. very little error) in the case of a two parameter hypo-elastic model for soil. When a complex six parameter constitutive model for the soil was used (i.e. nonlinear elastic behavior with extended Drucker-Prager failure criteria and associated flow rule), this back analysis technique worked reasonably well. Although, some parameters were identified with very little error, other parameters contained up to about 25% error. Occasionally, the technique did not converge to a reasonable solution. Inclusion of a third-order correction improved the accuracy of the technique considerably. The technique was not very sensitive to the presence of random normal noise.