PROBABILITY ESTIMATION OF OCCURRENCE OF MAJOR EARTHQUAKES IN THE "UC" UNDERGROUND MINING USING PRINCIPAL-COMPONENT-ANALYSIS-BASED ARTIFICIAL NEURAL NETWORKS

• Main Author: Husein Shihab, Nabiel
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/51556
• Institution: Institut Teknologi Bandung

Mining activities can change the subsurface stress condition so that it can induce microseismics and can cause earthquakes with large magnitudes (major events). There are several studies on the potential for disasters due to earthquake induced using different attributes, such as log EI, apparent stress, apparent volume, b-value, seismic wave velocity, etc. In this study, the author estimated an area with a high probability of a major earthquake occurrence (Mw ? 0.7) using principal-component-analysis-based artificial neural network (ANN). The multi-attributes used are log EI, apparent stress, apparent volume, Es / Ep, number of earthquakes, b-value, seismic wave velocity (Vp, Vs, Vp / Vs), blasting effect, and lithology model of the study area. The study area will be divided into boxes with a size of 40 m x 40 m. Each box will have the value of each attribute and a probability will be determined for a significant earthquake to emerge. The multi-attribute is analyzed using PCA. Seven principal components that have represented 94% of the data variation are used as features for ANN. In this study the authors estimated the probability of a major earthquake occurrence in the next two weeks using data from the previous two weeks. From the results of ANN modeling, we show that in general there are two zones in the study area that have different ANN model performance. The two zones are the areas in the south and west of the cave with f1 scores of 17.02% and 47.42% respectively. The difference in the results might be caused by different lithology and mechanical properties in the two areas.