An investigation on the Su–NSPT correlation using GMDH type neural networks and genetic algorithms

The Standard Penetration Test (SPT) is perhaps one of the most effective tests for quick and inexpensive evaluation of the mechanical properties of soil layers. There have been numerous studies directed towards establishment of correction factors for SPT blow count (NSPT) and correlations between NSPT and the properties of cohesionless soils. However, the test method is commonly used in all types of soils. It is, therefore, necessary to investigate the applicability of the correction factors and develop the appropriate correlations for fine-grained soils. er to investigate the relevancy of the overburden correction factor for NSPT in fine-grained soils, as well as establishing a correlation between undrained shear strength of such soils with NSPT, a data bank of SPT results on low plasticity fine-grained soils has been compiled. The effect of natural moisture content, plasticity index and effective overburden stress on the correlation of SPT–N60 and undrained shear strength of the soils has been studied by the use of Group Method of Data Handling (GMDH) type neural network optimized with genetic algorithms (GA). h this study a correlation has been obtained, expressing undrained shear strength of low-plasticity (PI < 20) fine-grained soils in terms of SPT–N60, PI and effective overburden stress. It has also been shown that natural moisture content has negligible effect on the correlation. The performance of this correlation was compared with other available correlations for this type of soil, and it has been shown that appreciable improvement in prediction of the output has been achieved.