1. Prediction of blast-hole utilization rate using structured nonlinear support vector machine combined with optimization algorithms.
- Author
-
Yu, Bingbing, Wang, Bo, Li, Yi, Zhang, Yuantong, and Wang, Guohao
- Subjects
OPTIMIZATION algorithms ,STANDARD deviations ,BLAST effect ,RECEIVER operating characteristic curves ,SUPPORT vector machines - Abstract
Blasting is the primary method for ultra-deep roadway engineering, which is facing the challenge of low footage caused by unsatisfactory blasting effects. Among all the evaluation indicators, the blast-hole utilization rate is the most important index for measuring blasting effect. Consequently, accurately predicting this index is essential for improving roadway excavation efficiency. In recent decades, the field applications of artificial intelligence have emerged as the prime method, yet the issue of data loss and large errors in large-scale data processing remains unresolved. In this study, novel Structured Nonlinear Support Vector Machine (SNSVM) is introduced as the primary research tool. To enhance prediction performance and accuracy, Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Sparrow Search Algorithm (SSA) are utilized to optimize the hyperparameters of SNSVM. The prediction models comprise fourteen influencing factors, constituting the comprehensive blasting effect prediction system based on artificial intelligence. The principal criteria for assessing the performance of various models are the error correlation coefficients (Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), R-Square (R
2 )) and the Receiver Operating Characteristic (ROC) curve (standard deviation rate (γ)). Among the models considered, SSA-SNSVM exhibited the greatest capability when the swarm size is 90. The RMSE, MAPE and R2 values of training datasets are 0.0070, 15.54% and 0.9295, respectively. The RMSE, MAPE and R2 values of testing datasets are 0.0086, 16.37% and 0.9490, respectively. Furthermore, the minimum standard deviation rate of SSA-SNSVM serves as the vital index for measuring the accuracy, with a value of 0.11. Subsequently, the sensitivity analysis results indicate that the most sensitive factor of blast-hole utilization rate is the surrounding rock itself. The comprehensive blasting effect evaluation is of significant importance for the dynamic adjustment of on-site blasting schemes, including roadway excavation, shaft excavation, or pressure-relief engineering. [ABSTRACT FROM AUTHOR]- Published
- 2024
- Full Text
- View/download PDF