The Random Finite Element Method, Its Implementation in Geotechnical Software Through Python, and a Comparison with the Random Limit Equilibrium Method

With increasing data and computational power, it is possible to take advantage of more sophisticated probabilistic tools in order to undertake a reliability-based design. Existing software packages can be extended with Python scripting to implement these methods, either through a scripting interface or through manipulating text file-based models. This paper demonstrates a digital tool that has been developed for the FEM software package RS2 which implements a variety of sensitivity analyses through parameter value manipulation. This ranges from varying a single parameter value of a homogenous, uniform material, to automated spatial parameter variation using the Random Finite Element Method (RFEM). RFEM involves the analysis of randomly generated, spatially variable virtual soils (volumes of soil parameters), within a Monte Carlo framework. It can be used for various statistical tasks, including estimating the probabilistic sensitivity of a design to a critical parameter, which can assist in reliability-based design. The use of random fields also allows for the risk of local inhomogeneity to be accounted for. The Python-based tool has a full graphical user interface to handle user inputs, and provides outputs as a set of useful tables and graphs. The steps involved in the tool are outlined, and an example scenario is presented for a slope stability problem with a building at its crest, and this RFEM scenario is compared to the Random Limit Equilibrium Method (RLEM). The implications of different statistical inputs are demonstrated in the results, along with the choice of RLEM versus RFEM. In addition, this paper aims to be a broad introduction to RFEM, with information from the availability of software to recommended parameters.