Evaluation of Creep Behavior of Soft Soils by Utilizing Multisensor Data Combined with Machine Learning

Area of Science:

• Geotechnical Engineering
• Computational Geomechanics
• Remote Sensing

Background:

• Burger's constitutive law is crucial for analyzing soft soil creep behavior.
• Traditional laboratory tests for creep parameters are costly and time-consuming.
• Accurate long-term behavior prediction is vital for geotechnical structures.

Purpose of the Study:

• To develop and validate a multisensor data-driven methodology for identifying Burger's creep parameters in soft soils.
• To overcome limitations of conventional laboratory testing methods.
• To provide reliable estimations of soft soil long-term behavior.

Main Methods:

• Integration of geophysical, geotechnical, and unmanned aerial vehicle (UAV) data for numerical modeling.
• Utilizing a custom-architecture neural network to analyze complex soil creep characteristics and displacements.
• Employing InSAR and GPS monitoring data with a particle swarm algorithm to determine optimal Burger's creep parameters.

Main Results:

• The proposed methodology successfully identified Burger's creep parameters using multisensor data.
• The method provides reliable estimations of the long-term behavior of soft soils.
• Validation on the Oostmolendijk embankment confirmed the methodology's effectiveness.

Conclusions:

• Multisensor data integration offers a cost-effective and efficient alternative to laboratory tests for determining soft soil creep parameters.
• The developed approach enhances the prediction accuracy of long-term geotechnical structure performance.
• This methodology is applicable to soft soil structures, improving engineering assessments.