The Porosity Estimation by Using an Adaptive Neuro-fuzzy Model from Well-log Data

Porosity is one of fundamental rock properties which relates to amount of fluid contained in reservoirs. In uncored intervals and wells to heterogeneous formations, porosity estimation from conventional well logs will be a difficult and complex task by statistical approaches. Intelligent computing approaches have been successfully used for rock characterization. It has been proved that the combined approaches such as neuro-fuzzy lead to better results. We used an Adaptive Neuro-Fuzzy Inference System (ANFIS) approach for estimating porosity. ANFIS is a neuro-fuzzy model that implements a Takagi-Sugeno fuzzy inference system and has a five layered architecture. The first hidden layer is for fuzzification of the input variables while T-norm operators are deployed in the second hidden layer to compute the rule antecedent part. The third hidden layer normalizes the rule strengths and is followed by the fourth hidden layer where the consequent parameters of the rule are determined. The output layer computes the overall output as the summation of all incoming signals. ANFIS uses backpropagation learning to determine premise parameters, and least-mean square estimation to determine the consequent parameters. In the study, one well was used for training and checking and another well for testing.