Evaluating the RMR correlation with the rock mass wave velocity using the meta-heuristics algorithms

Area of Science:

• Geotechnical Engineering
• Rock Mechanics
• Applied Geophysics

Background:

• Rock Mass Rating (RMR) is crucial for rock engineering design.
• Traditional RMR determination is destructive, costly, and time-consuming.
• Non-destructive methods using easily measurable parameters offer an economic alternative.

Purpose of the Study:

• To evaluate the relationships between RMR and seismic wave velocities (Vp and Vs).
• To develop and compare non-destructive models for RMR prediction.
• To identify the most accurate predictive model for RMR using Vp and Vs.

Main Methods:

• Analysis of 150 in-situ datasets across diverse rock types.
• Application of Genetic Algorithm (GA), Trust Region Reflective (TRR), and a hybrid TRR-GA model.
• Validation using Random Forest (RF) analysis and various performance metrics (scatter plots, error histograms, Taylor diagrams, RER curves).

Main Results:

• Simultaneous use of Vp and Vs is more reliable for RMR determination than individual parameters.
• All proposed models (GA, TRR, TRR-GA) demonstrated high accuracy in RMR prediction.
• The hybrid TRR-GA model exhibited superior performance in predicting RMR based on Vp and Vs.

Conclusions:

• The hybrid TRR-GA model offers a robust, non-destructive, and accurate method for RMR determination.
• Utilizing cost-effective seismic velocities (Vp, Vs) with meta-heuristic algorithms enhances RMR assessment.
• Further validation with larger datasets and diverse rock types is recommended for practical application.