Soil Deposit Stochastic Settlement Simulation Using an Improved Autocorrelation Model

Abstract

Abstract Soil parameters are spatially random variables. Thus, the spatial correlation relationship, besides the mean and variance, of a specific soil site is needed for any realistic stochastic modeling. In this regard, an improved autocorrelation model involving a linear, an exponential and cosine terms, named linear-exponential-cosine (LNCS), is adopted here to capture the spatial properties of the soil deposits. Further, a random field of the soil deposit is simulated using a two-dimensional Karhunen–Loeve expansion based on the new autocorrelation model. Furthermore, two cases for the soil settlement are calculated with the random field of the soil deposits. One case is the stochastic settlement from a reference paper. Some comparisons are undertaken, and it is found that the mean value agrees well with the reference. The other case involves the differential settlement analysis of a real engineering project. The settlement is calculated with the random field, the uniform field respectively, and is compared with the on-site measured values. The results show that the random field model can capture the differential settlement better than the corresponding uniform field model.