Application of Fuzzy System with Deep Learning in Seismic Facies Analysis

Summary Recent progress in deep learning, especially convolutional neural networks, has brought new advances in the automatic interpretation of seismic data. However, limited by the properties of seismic datasets itself and the characteristics of interpretation tasks, the generation of massive training samples, data noise and label uncertainty become the main technical bottlenecks. Taking the seismic facies identification as an example, in order to solve the problems of the low signal-to-noise ratio and inaccurate labels, a hybrid model of fuzzy system and deep neural network is proposed. Experiments show that the hybrid model provides the better classifications of facies near the boundary between geologic units and is robust against to noise due to the introduction of the fuzzy rules simulating the thinking mode of the interpreter. The application of Adaptive Network-Based Fuzzy Inference (ANFIS) improves the interpretation accuracy of seismic facies classification.