A framework for determining the optimal moisture content of high-speed railway-graded aggregate materials based on the lab vibration compaction method

Abstract

The Lab Vibration Compaction Method (LVCM) has been demonstrated to be the most effective method for compacting High-Speed Railway Graded Aggregate (HRGA). The Optimal Moisture Content (OMC) of LVCM is traditionally determined using the Heavy Hammer Compaction Method (HHCM). However, the differing compaction mechanisms of these two methods raise questions regarding the applicability of the OMC determined by HHCM. To address this issue, a novel assessment method was developed to evaluate the compaction quality of specimens in LVCM from both physical and mechanical perspectives (e.g., ρdn, Krb, and K20). Through a series of vibration compaction tests, the physical and mechanical properties of HRGA specimens were explored for various moisture contents in order to define the OMC. Additionally, three Hybrid Machine Learning (ML) models (PSO-ANN, PSO-SVM, and PSO-RF) were trained using experimentally obtained data and relevant HRGA material features to predict the OMC. The results indicate that the Critical Moisture Content (CMC) index can serve as a standard criterion for determining the OMC in LVCM and that HRGA materials with CMC possess desirable strength and load-bearing capacity. Furthermore, the experimental data suggested that graded and water abrasion features of HRGA materials significantly affect the CMC. The PSO-ANN model has the highest predictive accuracy with R2 = 0.94, MSE = 0.141, and MAE = 0.28 and the lowest uncertainty with U95 = 1.028 and Tstat = 2.816 compared to other hybrid ML models. The findings of this study hold significant implications for advancing the widespread implementation and utilization of the Lab Vibration Compaction Method (LVCM).