A Statistical Investigation of Dynamic Cone Penetrometer Test

Despite the widespread application of the dynamic cone penetrometer (DCP) test in geotechnical engineering, few empirical equations were previously introduced for relating the dynamic cone penetrometer index (DCPI) and other factors based on a very limited set of data. This paper aimed to systematically study all key factors affecting the DCPI number in different soil types. Thus, many in-situ DCP tests were performed in different depths and locations in which the ground was comprised of gravel, sand, silt, and clay soils with wide range of moistures and densities. Then, a comprehensive database of DCP tests involving the test results herein and other tests from the literature was statistically processed and an artificial neural network (ANN) model is proposed for predicting DCPI of soil. While the previous equations of DCPI were developed based on limited numbers of input variables including moisture content and density, the proposed ANN model involves many input parameters including depth of test location, gravel and fine contents of the soil, liquid limit (LL), plasticity index (PI), moisture content, and dry density. Based on the sensitivity analyses, the PI and moisture content have the greatest impact on the DCPI of soil, and the gravel content and depth of test location have the lowest importance among the input variables of the ANN. According to the parametric analysis, the DCPI of the soils decreases with increasing the test depth, dry density, and in turn, increases with an increase in the moisture content, LL, and PI of the fine fraction of soil. The DCPI of silt and clay soil is strongly affected by any change in the input variables.