# Assessment of liquefaction resistance of soil with fines using cyclic hollow cylinder testing

## TL;DR

This paper studies how the presence of fines in soil affects its resistance to liquefaction during earthquakes using laboratory tests and a neural network model.

## Contribution

The study introduces a new finding that increasing fines content beyond 15% enhances soil liquefaction resistance and uses a BPNN model for accurate predictions.

## Key findings

- Increasing fines content up to 15% reduces liquefaction resistance, but beyond 15% it increases resistance.
- A BPNN model predicted cyclic stress ratios with high precision (MAPE of 1.05%).
- Cyclic hollow cylinder tests revealed the influence of fines content, confining pressure, and relative density on liquefaction behavior.

## Abstract

Soil liquefaction is a devastating effect of earthquakes. It occurs when saturated granular soils lose their shear strength because of a sudden increase in pore water pressure under dynamic loads. Over the last six decades, considerable focus has been placed on understanding the mechanisms and phenomena associated with liquefaction, making it a critical area of research. Evaluating soil liquefaction accurately is crucial for maintaining the seismic safety of construction. To investigate how fines content affects soil liquefaction resistance under cyclic simple shear loads and gradual principal stress rotation, a series of cyclic hollow cylinder tests (CHCT) were carried out under isotropic consolidation and undrained conditions. The experiments were conducted using medium Monterey No. 0/30 Sand (MS), where five varying percentages of fine content were analyzed under two confining pressures (σ3’ = 103 kPa and 206 kPa) and at two relative densities (Dr = 30%, 45% and 60%). The results of CHCT tests contributed to the development of liquefaction-potential evaluation curves. The findings demonstrated that increasing the acceptable fines content up to 15% reduces liquefaction resistance. However, when fines content exceeds 15%, further increases lead to enhanced liquefaction resistance. Based on all laboratory test results, back propagation neural network (BPNN) was applied to predict cyclic stress ratios leading to initial liquefaction after cyclic loading cycles. The BPNN model can give superior precision with mean absolute percentage error (MAPE) values of 1.05%, and also can help engineering better understand liquefaction potential of soil samples with different fines content.