Predictive correlations for the compaction of clean sands

This paper presents a summary of a laboratory investigation on the compaction of clean sands and a method to predict the density of material as a function of the applied energy. The experiments, carried out on marine quarzitic-calcareous sand systematically assorted to obtain sixteen different grain size distributions, include the determination of minimum and maximum index dry density together with Proctor compaction tests performed with different levels of energy. The outcomes of this experimentation have then been merged with the results of several previous works to form an extensive database used to validate the correlation among the different variables. After an overview of the literature to identify the fundamental factors of compaction and examine the existing correlations, it is suggested to synthetically express the role of all the inherent properties of the soil aggregate i.e. grading, shape and roughness of grain, with the maximum index void ratio (emax). The efficacy of this choice is confirmed by the straight dependency on emax of the compactability of the material, defined as the difference between maximum and minimum index void ratios (emax - emin), and of the void ratios obtained with Proctor tests at fixed levels of energy. In particular, the combined role of emax and compaction energy is quantified by an artificial neural network to formulate a predictive chart giving errors lower than 10%.