LITHOLOGY DISCRIMINATION BASED ON MACHINE LEARNING

Abstract

Lithology discrimination (i.e., lithofacies classification) is significantly crucial in seismic interpretation. Comparing with time-consuming, labor-intensive and experience-dependent traditional method, the new method can predict the lithology using well log data efficiently and do not need the experience from domain experts. Though the study, it is concluded that the new method presented in this research is reliable and enough to predict generally the lithology using sufficient well log data from different locations and different environments. The data processing strategies play an essential role in improving the reliability of the model.