Open AccessEstimating pile length uncertainty with Kriging
Author(s) -
Martti Hallipelto,
Ari Hartikainen,
Taavi Dettenborn
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/710/1/012074
Subject(s) - kriging , probabilistic logic , pile , borehole , process (computing) , uncertainty analysis , computer science , uncertainty quantification , pipeline (software) , variance (accounting) , statistical model , engineering design process , propagation of uncertainty , mathematical optimization , engineering , algorithm , simulation , machine learning , artificial intelligence , geotechnical engineering , mathematics , mechanical engineering , accounting , business , programming language , operating system
Piles are used to transfer loads from the structure into the bearing soil layer. The accurate estimation of the pile lengths is a complex statistical problem that includes several spatial and measurement-based uncertainties. Commonly, bearing subsurface is modeled with linear triangular networks, and geometry related uncertainty is considered with overall safety factor. Currently, it is common practice is to solve the unknowns in this problem with simplistic borehole analysis and engineering judgment. Our paper presents a statistical pile length model based on the borehole uncertainty analysis. Modern geotechnical designing contains multiple steps including ground investigations, computer modeling, and expert decision making. All the included steps add uncertainty to the design process. The problem with these traditional methods is that the sources and the magnitudes of the uncertainty are not visible to the designer, which can lead to non-optimal decisions. With the help of the state-of-the-art probabilistic models, a design pipeline can be constructed that propagates the uncertainty from process to process and is transparent about its sources and magnitudes. In our paper, we show how the probabilistic pipeline approach enables us to make more informed decisions. This is due to uncertainty propagation from ground investigations to the 3D volumetric model. Given the uncertainty, we can estimate the mean and variance of the future costs and the impact that different design decisions have. The paper presents a practical example where probabilistic models are utilized to improve decision making.