Probabilistic Prediction of Reinforcement Loads for Steel MSE Walls Using a Response Surface Method

The response surface method (RSM) with a quadratic polynomial was used to formulate three equations for the calculation of the maximum reinforcement loads in steel-reinforced mechanically stabilized earth (MSE) walls under operational (working stress) conditions. The RSM models were formulated using control variables found in the simplified stiffness method. The model coefficients were back-calculated from a large database of measured steel-reinforcement loads from full-scale instrumented walls using a least-squares solution. Model uncertainty was quantified using bias statistics for which model bias is defined as the ratio of measured to calculated reinforcement load. The simplest of the three RSM models has only three empirical constants and has the advantage that reinforcement stiffness and soil friction angle are not required as input parameters. The same model was shown to give a predicted load accuracy that exceeds that of the simplified method that is used in current U.S. design specifications and has the same practical accuracy as the simplified stiffness method used for steel MSE walls constructed with frictional soils. The paper shows how the three models can be used in Monte Carlo simulations to compute probabilities of load exceedance at the time of design.