Optimized adaptive neuro-fuzzy inference system for predicting blast-induced ground vibration in quarries based on hunger games search optimization

ABSTRACT

This paper aims at developing an optimisation model based on the adaptive neuro-fuzzy inference system (ANFIS) and hunger games search (HGS) algorithm for predicting the intensity of ground vibration in quarries, abbreviated as HGS-ANFIS. The stochastic optimisation characteristics of the HGS were considered and used to optimise the ANFIS model for this aim. The blasting parameters of 126 blasting events were investigated in this study. Several benchmark models, such as ANFIS (without optimisation), random forest (RF), and support vector machine (SVM), were also considered to compare and demonstrate the superiority of the HGS-ANFIS model. They were then applied in practical engineering through five blasts to in-situ validate their accuracies. The results showed that the improved ANFIS model by the HGS optimiser yielded the most dominant accuracy in this study with an RMSE of 0.274, R² of 0.985, and MAE of 0.216. The validation results in practical engineering also indicated that the accuracy of the HGS-ANFIS model improved by 2% compared with the ANFIS model without optimisation and 4 —12% compared with the RF and SVM models in practical engineering.

KEYWORDS:

• Artificial intelligence
• ANFIS
• hunger games search optimisation
• ground vibration
• rock blasting
• peak particle velocity

Acknowledgments

This research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) under grant number 105.99-2019.309. The authors would like to thank the engineers in the Thuong Tan 3 quarry, Vietnam to support and provide the details of the study site and dataset.

Disclosure statement

The authors have no known conflicts of interest to declare associated with this publication.