DENSITY LOG PREDICTION BY USING AN IMPROVED BP NEURAL NETWORK
Lingjun Chen · Computing Techniques for Geophysical and Geochemical Exploration · 2007
One of important works of seismologist is the integration of well-log and seismic data. We can derive the relationship between seismic data and well-log by analyzing training data at the position of well. The statistical relation will be constructed by extracting seismic data attributes to predict logs. This paper chooses an improved BP neural network with a three-layers structure. An example shows that the features of the network algorithm are of rapid speed, mathematical simplicity and ability of avoiding local minima. Applying this network to one 2-D seismic data section at practice,density log is predicted successfully by constructing the nonlinear relation between extracted seismic multi-attribute and log, which greatly benefit our understanding of the reservoirs.