Predicting ground vibration during rock blasting using relevance vector machine improved with dual kernels and...

Area of Science:

• Geotechnical Engineering
• Computational Intelligence
• Environmental Monitoring

Background:

• Ground vibration from rock blasting poses significant environmental and safety risks.
• Accurate prediction of peak particle velocity (PPV) is crucial for mitigating these hazards.
• Existing methods require robust predictive models for effective risk assessment.

Purpose of the Study:

• To develop and compare novel Relevance Vector Machine (RVM) models for predicting PPV in quarry blasting.
• To identify the most efficient RVM model for ground vibration estimation.
• To provide a tool for engineers to select optimal parameters for vibration prediction.

Main Methods:

• Employed conventional and optimized RVM models for PPV prediction.
• Utilized a particle swarm optimization (PSO) approach to enhance RVM performance (PSO_DRVM).
• Evaluated thirty-three RVM models, including the proposed PSO_DRVM model MD29, using various performance metrics.

Main Results:

• All tested RVM models achieved a performance score above 0.85, indicating strong prediction accuracy.
• The PSO_DRVM model MD29 exhibited superior performance compared to other RVM models.
• Key performance metrics included RMSE of 16.2272 mm/s, R-value of 0.9175, and IOA of 0.8239.

Conclusions:

• Optimized RVM models, particularly PSO_DRVM MD29, are highly effective for predicting ground vibration from blasting.
• The study provides valuable insights for selecting appropriate kernel functions and hyperparameters.
• This research can significantly enhance safety protocols and operational efficiency in the mining and civil industries.