Feature Importance of Stabilised Rammed Earth Components Affecting the Compressive Strength Calculated with...

Area of Science:

• Construction Materials Science
• Sustainable Engineering
• Geotechnical Engineering

Background:

• Cement-stabilized rammed earth (CSRE) is a sustainable and economical building material.
• The variable nature of local soils used in CSRE mixtures leads to inconsistent compressive strength.
• Lack of standardized mixture designs hinders widespread adoption and reliable performance prediction.

Purpose of the Study:

• To identify the key components influencing the compressive strength of CSRE.
• To develop predictive models for CSRE compressive strength using machine learning.
• To provide a basis for optimizing CSRE mixture design for enhanced performance.

Main Methods:

• Utilized a database of 434 CSRE samples for analysis.
• Employed machine learning regression models: artificial neural networks, decision tree, and random forest.
• Applied explainable artificial intelligence (XAI) tools to rank ingredient influence on compressive strength.

Main Results:

• Machine learning models demonstrated low prediction errors for CSRE compressive strength.
• XAI tools consistently identified the most influential components affecting compressive strength.
• Results were validated through statistical analysis of selected features.

Conclusions:

• The study successfully identified critical factors governing CSRE compressive strength.
• The developed approach provides a reliable method for optimizing CSRE mixture design.
• This research supports the advancement of sustainable construction materials through data-driven insights.