Developing soft-computing regression model for predicting soil bearing capacity using soil index properties

There are challenges associated with determining the actual behaviour of soil foundation due to its heterogeneous nature especially when subjected to an imposed loading. This has led to deployment of various soft-computing approaches as an alternative to field experiments for measuring the resistance level of the soil caused by imposed loading. Therefore, this research work was aimed at modelling the soil bearing capacity with specific consideration of index properties parameters of soil, shear strength parameters and relative varied depths, by employing Terzaghi's equations. To considerably overcome complexities, and spontaneous variabilities associated with natural soil foundation, a relatively larger set (45) of data were sourced and used for the modelling. Multiple linear regression (MLR) was used to develop the model using 30 set of data and natural bearing capacity was determined as output with relatively high level of accuracy. The model developed was validated using remaining 15 set of data by employing normal probability plots, which indicates a low level of variance between experimental, and modelled values. Likewise, the corresponding R-2 values of strip, square, and circular footings were found to be 96.98%, 96.93%, and 96.90%, this indicates a high reliability of the model developed.