Automated intelligent hybrid computing schemes to predict blasting induced ground vibration

Blasting has been widely recognized as an economical and viable method in geo-engineering projects. However, the induced ground vibration in terms of peak particle velocity (PPV) potentially can damage the nearby environment and inhabitants. Therefore, more accurate prediction of the PPV can lead to reduce undesirable and hazardous effects of blasting. With the increase in the computational power, wide variety of predictive PPV models using numerical tools and data mining approaches have been presented. In this paper, the optimum predictive PPV model was specified using generalized feedforward neural network (GFFN) structure integrated with a novel automated intelligent setting parameter approach. Subsequently, two new optimized hybrid models using GFFN incorporated with firefly and imperialist competitive metaheuristic algorithms (FMA and ICA) were developed and applied on 78 monitored events in Alvand-Qoly mine, Iran. According to analyzed metrics, the predictability level of hybrid GFFN-FMA dedicated 6.67% and 20% progress than GFFN-ICA and optimum GFFN. The pursued performance using precision-recall curves and ranked accuracy criteria also exhibited superior improvement in GFFN-FMA. Sensitivity analyses implied on the importance of the distance and burden as the most and least effective factors on predicted induced PPV in the study area.

Automated summary

The study established an optimal predictive PPV model using a generalized feedforward neural network structure and a novel automated intelligent parameter setting approach. Two new optimized hybrid models were developed, incorporating GFFN with firefly and imperialist competitive metaheuristic algorithms (FMA and ICA). Results showed that the predictability level of the hybrid GFFN-FMA model significantly outperformed the GFFN-ICA and optimum GFFN models.