MULTIPLE ATENUATION IN SHOT GATHER BY USING CONVOLUTIONAL NEURAL NETWORK (CNN)

Today Machine Learning is used in almost every field for human life, including geophysics. Some examples of Machine Learning utilities are classifying lithology and predict petro physical parameters based on several supported data. Especially in seismic method, Machine Learning can be used for removing or attenuate multiple from seismic image or shot gather data by using Convolutional Neural Network (CNN). It reduces the multiple from shot gather data (input) based on filtered shot gather data (called by ground truth model) as the label or target. Unfortunately, filtering process sometimes erase boundaries layer in shot gather. Then CNN works by generating several activation function in neurons and hidden layers, multiply with input data and reconcile them to labels to reinforce the boundaries. To validate the CNN result, it can be seen from L – curve as the loss function that represent the prediction error. The fewer the prediction error, the more accurate result is observed.