Researchers have released a comprehensive corpus of 2,854 laboratory soil compaction tests, sourced from six public datasets and spanning 162 provenance groups and four Proctor energy levels. This dataset addresses limitations in existing correlations, which are often based on fewer specimens and a single compactive energy. The corpus includes detailed information on fines content, zero-air-voids conditions, and specific gravity, revealing that a significant portion of published compaction data is physically impossible. Machine learning models, including a tabular foundation model and symbolic regression, were developed to predict density and water content, achieving notable R2 scores, particularly when accounting for compactive energy and provenance. AI
IMPACT Provides a large, open dataset and ML models for soil compaction analysis, potentially improving engineering accuracy.
RANK_REASON The item is an academic paper detailing a new dataset and machine learning baselines for soil compaction tests. [lever_c_demoted from research: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
