Mine Induced Airblast prediction: An Application of Chaos Game Optimization based soft computing approaches

Abstract

Abstract Air overpressure, often known as AOp, is one of the unfavourable effects of galena blasting. It has high damage potential to structural buildings, ecosystem, and occasionally endanger mine workers due to the ore characteristics. The efficiency of most available techniques to manage this mine challenge is site specific and sometimes limited by poor prediction accuracy. In the current study, several deep and machine learning approaches have used to develop blast-induced AOP prediction models as a way forward to the recent gap. These techniques include, Long Short-Term Memory (LSTM), Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), Random Forest (RF), and Multivariate Adaptive Regression Splines (MARS). To achieve the research aim, 90 measured field data sets were monitored and collected from the Anguran open-pit lead-and-zinc mine (Iran). To improve the proposed model performance, chaos game optimisation (CGO) as a nature-inspired algorithm was adopted. The optimized models were compared with each other and as well with the ordinary models to determine the model with the best prediction accuracy. The results show that, in terms of accuracy levels, the performance of hybrid algorithm approach is superior to that of single based models. The LSTM-CGO model, out of the 10 proposed models, has the highest prediction accuracy statistically. This study demonstrated how well deep learning techniques provide solution to safe and environmental friendly mining operation. The proposed soft computing models are applicable as a tool to forecast AOP in surface mine blasting operation as a pre-blast design decision making reference.