A PSO-XGBoost Model for Predicting the Compressive Strength of Cement–Soil Mixing Pile Considering Field Environment Simulation

Cement–Soil Mixing (CSM) Pile is an important technology for soft ground reinforcement, and its as-formed compressive strength directly affects engineering design and construction quality. To address the significant discrepancy between laboratory-tested strength and field as-formed strength arising...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiagui Xiong, Yangqing Gong, Xianghua Liu, Yan Li, Liangjie Chen, Cheng Liao, Chaochao Zhang
Format: Article
Language:English
Published: MDPI AG 2025-08-01
Series:Buildings
Subjects:
Online Access:https://www.mdpi.com/2075-5309/15/15/2740
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Cement–Soil Mixing (CSM) Pile is an important technology for soft ground reinforcement, and its as-formed compressive strength directly affects engineering design and construction quality. To address the significant discrepancy between laboratory-tested strength and field as-formed strength arising from differing environmental conditions, this study conducted modified laboratory experiments simulating key field formation characteristics. A cement–soil preparation system considering actual immersion conditions was established, based on controlling the initial water content state of the foundation soil before pile formation and applying submerged conditions post-formation. Utilizing data mining on 84 sets of experimental data with various preparation parameter combinations, a prediction model for the as-formed strength of CSM Pile was developed based on the Particle Swarm Optimization-Extreme Gradient Boosting (PSO-XGBoost) algorithm. Engineering validation demonstrated that the model achieved an RMSE of 0.138, an MAE of 0.112, and an R<sup>2</sup> of 0.961. It effectively addresses the issue of large prediction deviations caused by insufficient environmental simulation in traditional mix proportion tests. The research findings establish a quantitative relationship between as-formed strength and preparation parameters, providing an effective experimental improvement and strength prediction method for the engineering design of CSM Pile.
ISSN:2075-5309