PREDICTION OF BLASTING FRAGMENTATION RESULT USING ANFIS METHOD

• Main Author: P Purba, Albert
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/54437
• Institution: Institut Teknologi Bandung

One of the indicators of blasting success is the distribution of fragmentation. The size and distribution of the blasted material will greatly affect the excavation, loading and processing stages. All that factors has influence and affect the cost of mining activities. Therefore we need a method which is good and reliable to predict the size of fragmentation. So far we have been known some methods to predict size and distribution of fragmentation which are quite helpful and commonly used. However, there are still weaknesses that make the prediction results unsatisfactory. This is because the system working in the blasting process is very complex and includes a wide range of parameters. Therefore, this study aims to apply the ANFIS (adaptive neuro-fuzzy inference system) method to predict blasting fragmentation then analyze and compare the results with the results of the MRA (multiple regression analysis) method. We hope the ANFIS method produces satisfactory predictions compared to other methods. The primary data is come from photos of the blasted material and the blast design as secondary one. This research used 110 data with 4 input variables, which is B / S ratio, depth, stemming, and PF. All fileld data collected at the PT Berau Coal site Lati (LMO) at the PQRT pit. Data photo processing tools use Split Desktop 2.0. The MRA processing model uses SPSS 20 and the ANFIS model uses MatLab 2018. Model analysis uses RMSE and VAF values. The result of RMSE values for the MRA models P20, P50, P80 and Topsize training were 0.009, 0.0317, 0.2458 and 0.1289, respectively. While the ANFIS model is 0.072, 0.0266, 0.0509 and 0.1107. The VAF results of MRA models P20, P50, P80 and Topsize training were 61.68, 65.08, 85.27 and 80.53, respectively. While the ANFIS method is 75.26, 79.08, 88.29, and 77.45, respectively. From the results above, it can be shown that the ANFIS method is able to work better than the MRA method and can be used as an alternative method for predicting fragmentation in addition to other commonly known methods.