We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Soft computing of the recompression index of fine-grained soils.
- Authors
Alzabeebee, Saif; Alshkane, Younis M.; Al-Taie, Abbas J.; Rashed, Kamal Ahmad
- Abstract
Consolidation settlement is a phenomenon happens in saturated fine-grained soils when subjected to change in effective stress. Consolidation settlement is often determined using the compressibility parameters, the compression index ( Cc ) and the recompression index (C r ). However, there is lack of studies on the accuracy of the empirical correlations to predict the recompression index ( Cr ). In addition, no study has been concerned with the development of region-specific correlations to predict the recompression index of fine-grained soils in middle and north of Iraq. Thus, this research has been conducted to fill these gaps by collecting and testing of 350 undisturbed samples, assessing the available empirical correlations and developing of novel region-specific data-driven correlations to predict the recompression index (C r ) using the evolutionary polynomial regression analysis (EPR-MOGA). The tests include the soil unit weight, specific gravity, water content, Atterberg limits and recompression index. The statistical assessment involved calculating the mean absolute error ( MAE ), root mean square error ( RMSE ), mean ( μ) , percentage of predictions within error range of ± 30% ( P 30 ) and coefficient of correlation (R ). In addition, 80% of the data has been used in the EPR-MOGA model training while 20% of data has been used to test (validate) the model accuracy. The results illustrate that all of the available correlations provide an average prediction of the recompression index with R ranges between 0.12 and 0.70, MAE ranges between 0.01 and 0.03, RMSE ranges between 0.020 and 0.030, μ ranges between 0.26 and 1.73 and % P 30 ranges between 0 and 61%. However, the new EPR-MOGA correlations showed much better prediction capabilities, where the best EPR-MOGA correlation scored R , MAE , RMSE , μ and P 30 of 0.90, 0.01, 0.008, 1.05 and 82%, respectively, for training data and 0.88, 0.01, 0.009, 1.02 and 86%, respectively, for testing data. The new EPR-MOGA correlations require only the saturated unit weight, water content and void ratio of the soil to accurately predict the Cr .
- Subjects
IRAQ; SOFT computing; STANDARD deviations; WATERLOGGING (Soils); SPECIFIC gravity; REGRESSION analysis
- Publication
Soft Computing - A Fusion of Foundations, Methodologies & Applications, 2021, Vol 25, Issue 24, p15297
- ISSN
1432-7643
- Publication type
Article
- DOI
10.1007/s00500-021-06123-3