Settlement Prediction – How Much Is It A Matter Of Luck?

Accurate prediction of settlement remains one of the most difficult tasks geotechnical engineers have to contend with. Typically, settlement prediction is undertaken using a deterministic approach. This involves the idealisation of a soil profile as one or several uniform layers each associated with a set of model parameters. This idealisation glosses over the spatial variations of parameters within each layer. Much of the hit and miss in settlement prediction is tied to the uncertainty or inadequate knowledge regarding the spatial variation and how to reasonably model these variations.

This paper attempts to quantify the amount of luck associated with a prediction. It is shown that whether the design uses the mean, or some other percentile value of the soil parameter, spatial soil variability renders the deterministic prediction incapable of providing any information about the likely error of the prediction. Accordingly, it is a question of luck should the prediction coincide with the observed settlement. In order to quantify the probability (or luck), the maximum settlement of a flexible circular footing was evaluated using elastic theory and Monte Carlo simulation of randomly variable Young’s modulus with depth.