Investigation of the links between railroad ballast gradation and the packing coordination number: a discrete element method and machine learning study
GRANULAR MATTER, cilt.29, ss.1-22, 2027 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 29
- Basım Tarihi: 2027
- Doi Numarası: 10.1007/s10035-026-01718-9
- Dergi Adı: GRANULAR MATTER
- Derginin Tarandığı İndeksler: Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Scopus, Materials Science & Engineering Collection (ProQuest), Pharma Collection (ProQuest), Technology Collection (ProQuest), Science Citation Index Expanded (SCI-EXPANDED), Compendex, INSPEC, zbMATH
- Sayfa Sayıları: ss.1-22
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Karadeniz Teknik Üniversitesi Adresli: Evet
Özet
The ballast layer, consisting of uniformly graded coarse aggregates, is paramount for load distribution from crossties to subgrade layers that often have low bearing capacity. The interaggregate contact network of a ballast layer plays a crucial role in this load distribution. A denser interaggregate contact network provides better load distribution, resulting in improved ballast performance, including higher shear strength and reduced degradation under loading. Consequently, establishing a dense contact network within a ballast layer is essential to improve its performance. The average number of contacts each aggregate has within a granular assembly, or coordination number (CN), is a key microscopic property that provides a quantifiable measure for contact network density. Gradation is one of the design elements that plays a definitive role in the magnitude of the CN. However, the understanding of the relationship between gradation and CN is lacking. This paper presents a discrete element method (DEM) packing study to remedy this shortcoming. The study incorporated various international specifications to create 867 gradations. CN values resulting from these gradations were recorded and analyzed using machine learning methods. The analyses indicated a weak inverse relationship between the coefficient of uniformity (Cu) and CN. The median particle size and the smallest particle size ratio displayed better single-predictor performance. An empirical equation to predict CN was developed using Elastic Net and Multiple Linear Regression. The components of the equation were further analyzed using XGBoost regression, and insights on the effects of gradation on CN were shared.