Prediction of recompression index using GMDH-type neural network based on geotechnical soil properties

Settlement based design for shallow foundations realizing the consolidation aspect is a major challenge in geotechnical engineering. The recompression index (Cr) from the oedometer test is used to estimate the consolidation settlement of over-consolidated (OC) clays. Since the determination of Cr from oedometer tests is relatively time-consuming and is usually determined for a single unloading, empirical equations based on index properties can be useful for settlement estimation. Correlations have been proposed to relate the Cr of clay deposits to other soil parameters. Since existing equations are incapable of estimating Cr well, artificial intelligence methods are used to predict them. In the present study, a Group Method of Data Handling (GMDH) type neural network is used to estimate the Cr from more simply determined index properties such as the liquid limit (LL) and initial void ratio (e0) as well as specific gravity (Gs). In order to assess the merits of the proposed approach, a database containing 344 data sets has been compiled from case histories via geotechnical investigation sites in the province of Mazandaran, along the southern shoreline of the Caspian Sea, Iran. In addition to the physical properties mentioned already, the natural water content (ωn), plastic index (PI) and dry density (γd) were also included in the model development. A comparison was carried out between the experimentally measured recompression indexes and the predictions in order to evaluate the performance of the GMDH neural network method. The results demonstrate that an improvement with respect to the other correlations has been achieved. Finally, a sensitivity analysis of the obtained model was performed to study the influence of the input parameters on the model output. The sensitivity analysis reveals that e0 and LL have significant influence on predicting Cr.