Assessment of the Relationship Between Rock Mass Classification Systems Using Hybrid XGBoost Algorithms: Insights from the Mechanized Tunnelling Project Database
Hanan Samadi,Jafar Hassanpour,Jamal Rostami
DOI: https://doi.org/10.1007/s00603-024-04227-4
IF: 6.518
2024-11-09
Rock Mechanics and Rock Engineering
Abstract:Rock mass classification systems play a crucial role in estimating the design parameters of rock engineering projects. However, the availability of geological data may sometimes be insufficient, rendering the utilization of a specific classification system impractical. Previous attempts by researchers to establish empirical correlations between classifications have tended to rely on simplistic regression analyses, limiting their applicability to specific geological conditions found only at the site of their development. Consequently, it is imperative to introduce geology-specific relationships empowered by hybrid intelligent predictor networks. To address this need, geological and engineering geological data from 107 sections across three mechanized tunnelling projects in diverse geological zones were collected. Subsequently, the data was categorized based on the particular rock mass conditions observed in the tunnels, and reasonable distribution intervals for the main factors, namely lithology and rock mass structure, were defined for each section. This study utilized a hybrid XGBoost methodology optimized with the salp swarm algorithm (SSA) and tree seed algorithm (TSA) to develop computational models predicting relationships between classifications like RMR, Q, BRMR, RCR, Q'c, N, and GSI under categorized engineering geological conditions. The effectiveness of these methods was assessed using a variety of loss functions and evaluation metrics, demonstrating their accurate prediction capabilities in linking rock mass classification parameters. Given the varied geo-mechanical properties and rock mechanical specifications of the hard rock tunneling projects in Iran, which encompass a range of lithological units, the outcomes and empirical formulas derived from these projects could be valuable for similar projects worldwide featuring comparable geological and rock mechanical characteristics. Consequently, these models offer potential for adoption within similar lithological units and geological zones.
geosciences, multidisciplinary,engineering, geological