The direct interpolation of a transfer function needs exponentially many data in terms
of the number of the fractions in the grading curve. The suggested transfer function
construction method - based on a double approximation technique, the grading entropy
concept and at most quadratic many data in terms of the fraction number - is tested
on the example of the dry density of sands here using some previously measured data.
In the first approximation step a "preliminary transfer function" is interpolated
in the non-normalized grading entropy diagram on the basis of some "optimal" soil
data. In the second approximation step the preliminary transfer function is extended
to the space of the possible grading curves with the constant function. The so determined
transfer function is tested against an independent "non-optimal" data set, measured
on some soil series with basically continuous (i.e., not gap-graded) grading curves.
The aim of this paper is to present the main results of the study supporting the goodness
of the method and the predictability of the dry density transfer function.