Random field generation of soil properties in geotechnical engineering considering the spatial correlation

Abstract

The use of random fields to describe soil parameters in geotechnical problems has become a popular trend in recent years. This approach is particularly essential in reliability analysis and risk assessment. Among the key aspects is the consideration of spatial correlation in random fields, which plays a crucial role. This study focuses on the Local Average Subdivision (LAS) method, which offers several advantages when applied to geotechnical problems involving spatial correlation. The research presents the theoretical foundation and algorithm of the LAS method and develops a computational program using the Python programming language. This program has been used to perform simulations and conduct the following comparisons: comparison with random fields that do not consider spatial correlation, which shows that the LAS model clearly demonstrates such correlation; comparison of different cases involving parameters such as variance and correlation length; and execution of 100,000 simulations to compare input results with expected values, variance, and correlation length. The results indicate that the generated model achieves an acceptable level of accuracy. However, the current program has only been applied to one-dimensional (1-D) problems, and the research team is continuing to develop it for application to higher-dimensional problems.