Machine learning-based method for determining regional parameters of the HSS model: a case study of Qingdao, China

Area of Science:

• Geotechnical Engineering
• Computational Mechanics
• Machine Learning

Background:

• The hardening soil small-strain (HSS) model accurately simulates complex soil stress paths but requires numerous parameters with significant regional variability.
• Determining these parameters is costly and time-consuming, hindering practical application in geotechnical engineering.

Purpose of the Study:

• To develop an efficient machine learning-based framework for determining locality-specific HSS parameters.
• To address the challenges of parameter determination cost and regional variability in geotechnical applications.

Main Methods:

• Statistical analysis of parameter ratios to identify sensitive parameters.
• Parameter sampling design for highly variable parameters.
• Training a BP neural network for nonlinear mapping and intelligent inverse analysis.
• Application to Qingdao, China, for plain fill, silty clay, and medium-fine sand.

Main Results:

• The machine learning framework successfully determined HSS parameters for typical Qingdao strata.
• Numerical simulations showed high consistency with on-site monitoring data (RMSE < 0.51 mm, MAE < 0.4 mm).
• Key inversion parameter ratios for Qingdao were determined for different soil types.

Conclusions:

• The proposed machine learning framework offers an efficient solution for determining site-specific HSS parameters.
• This approach enhances the practical applicability of advanced constitutive models in geotechnical engineering.
• The study provides crucial regional HSS parameter ratios for Qingdao, China.