COMPARISON OF PPV PREDICTION BETWEEN EMPIRIC METHODS AND ARTIFICIAL NEURAL NETWORK (ANN) FOR BLASTING OPERATIONS IN PARINGIN PIT, PT ADARO INDONESIA

• Main Author: Fitrabuana, Asril
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/61242
• Institution: Institut Teknologi Bandung

PT Bukit Makmur Mandiri Utama (BUMA) is one of the largest coal mining contractor in Indonesia and have many sector with different mining business permit (IUP) holders, which are located in Kideco, SDJ, Lati, Binungan and Paringin. One of leading companies in coal industry that are business partners with PT BUMA is PT Adaro Indonesia. In this study, a comparison of Peak Particle Velocity (PPV) predictions between empirical method and the artificial neural network was carried out. Three theories are used in the empirical method, that are USBM, Langefors-Kihlstorm and Ambrassey-Hendron. In the ANN method, seven input parameters are used that is diameter, burden, spacing, hole depth, holes, maximum charge and distance. In addition, it analyzes which blasting variables significantly affect the PPV, the aim is to find out the most optimal method to produce a PPV prediction and the more influential blasting variables to be taken into consideration in carrying out further blasting. To evaluate the results of the predictive models of the two methods is comparing the values of R2 and MAE, while to see the blasting variables that have significant effect on PPV, Multiple Linear Regression (MLR) analysis is used by looking for the significance value of each variables. In the empirical method the values of R2 and MAE respectively, USBM is 0,785 and 0,446; Langefors-kihlstorm is 0,351 and 0,517; and Ambrasseys-Hendron is 0,789 and 0,438. While the ANN method obtained a value of 0,896 and 0,267. In the MLR test, the correlation regression values of each parameters, namely diameter, burden, spacing, hole depth, holes, maximal charge and distance respectively are 0,567; 0,327; 0,225; -0,106; -4,11×10-4; 0,004; -0,002 and 0,002. Meanwhile, the significance value is 0,076; 0,058; 0,051; 0,683; 0,046; 2,41×10-23 and 0,567. Among the seven blasting variables used, that have a significant effect on the resulting PPV are the maximum charge weight and distance because they have a significance value of <0.05. In summary, the ANN is the best method in predicting PPV and the blasting variables that have significant effect on the resulting PPV are the maximum charge weight and distance.