Mining blast vibration forecasting based on a deep learning machine optimized by improved dung beetle optimization

The ground vibrations resulting from mining blasting operations represent the most considerable adverse impact on local inhabitants and the surrounding environment. Precisely forecasting blasting vibrations with a constrained amount of monitoring data is a feasible strategy to manage ground vibrations. This research introduced an innovative hybrid modeling approach that leverages the maximal information coefficient (MIC), deep extreme learning machine (DELM), and improved dung beetle optimization (IDBO) to predict both the peak particle velocity (PPV) and frequency. Initially, feature selection was conducted utilizing the MIC algorithm. Following this, the variables identified by the MIC were employed as inputs to construct the DELM model. To enhance the DELM model's performance, the IDBO was implemented to optimize the DELM model's hyperparameters. The findings from the experiment show that the maximum root mean squared error (RMSE), mean squared error (MSE), and R² of the proposed hybrid framework are only 0.237, 0.108, and 0.975, respectively. These outcomes indicate that the hybrid MIC-IDBO-DELM model holds great potential as a predictive tool for blasting vibration prediction.