As infrastructure development accelerates, ensuring the quality of the subbase layer in roadworks has become increasingly vital. Among various evaluation tools, the Dynamic Cone Penetration (DCP) test is widely recognized for its practical advantages—namely its ease of use, affordability, and ability to deliver real-time, continuous assessments of soil strength directly on-site without disturbing the ground. The research involved conducting both DCP and SRM tests on subbase materials classified as types B, C, and D, which are frequently utilized in Basra’s Road construction. The investigation measured parameters such as the Dynamic Cone Penetration Index (DCPI), moisture content, and dry density under three distinct moisture conditions, all assessed within a controlled laboratory setting. Results were analyzed using SPSS (version 27), revealing a strong inverse relationship between dry density and DCPI, A direct correlation between DCPI and moisture content and between moisture content and dry density. Three predictive equations were developed for each subbase type. The approach has proven to streamline testing processes by minimizing time and resource demands, making it a credible and efficient alternative to conventional subgrade resistance methods for field-based soil assessment.
Field dry density is a soil compaction characteristic that is useful for geotechnical engineering design. Laboratory methods are laborious and time-consuming. The purpose of this paper is to determine and compare the effectiveness of MLR and SVM models in predicting this essential parameter from the fundamental soil property index level. A dataset of 86 soil samples with various geotechnical qualities was used, containing data such as gravel, sand, fines, liquid limit, and plastic limit. The dataset was split into 80% training and 20% testing. Using $R^{2},$ RMSE, and MSE, the performance of the built MLR and SVM prediction models was thoroughly examined. With an R2 value of 0.988 (on the test set), the SVM model outperforms the MLR model in terms of prediction accuracy for FDD $(R2=0.814)$. Compaction behavior and soil property index properties have a complicated relationship, as seen by the performance gap. According to feature importance analysis, the SVM model's predictions heavily relied on the fines content. According to this study, SVM is a useful method that geotechnical engineers can employ to quickly and affordably estimate compaction parameters in the early phases of site investigations and design optimization.