Development of ANN-Based Universal Predictor for Prediction of Blast-Induced Vibration Indicators and its Performance Comparison with Existing Empirical Models

Air overpressure (AOp) and peak particle velocity (PPV) are the most undesirable effects of blasting in mines. It is an urgent need to design a predictor model for AOp and PPV for universal applications to minimize environmental effects and damages that occur due to blasting. The present study attempts to design an artificial neural network (ANN) model for the prediction of AOp and PPV in different conditions. In this study, the blast design parameters (number of holes, depth of the blast hole, stemming length, spacing, burden, distance of vibration monitoring location from the blast site, charge weight per delay) of two active mines (coal and iron ore) are considered in four different conditions for training and testing of the model for the prediction of AOp and PPV. The model was trained and tested in four different combinations of data (trained using data of one mine and tested using data of another mine, trained and tested using data of coal mine, trained and tested using data of an iron ore mine, trained and tested using combined data of both the mines) for examining the applicability in different conditions. The results indicate that R2 values are ranged from 0.886 to 0.908 and 0.8728 to 0.8959, respectively, in the prediction of PPV and AOp. A comparative study of the performances of the developed model with the other empirical model is also demonstrated. For this, the site constants of different empirical models were estimated individually for both the mines. The study results indicated that the ANN model performs much better than all the empirical models. It can be inferred from the results that blast vibration cannot be accurately predicted only from charge per delay and distance from the blast hole. The ANN model was considered many other factors and thus can predict the vibration level more accurately.