back-propagation neural network

Having information about Young’s modulus is extremely essential for characterization of the hydrocarbon reservoirs. This property can be conventionally determined by core sample data analysis in laboratory that is time-consuming, critically expensive and discontinuous. Therefore, many researchers have always been looking for suitable methods to estimate Young’s modulus with acceptable accuracy. The current research aims to create an advanced, precise model for estimating Young’s modulus by utilizing backpropagation neural network (BPNN), support vector regression (SVR), and gene expression programming (GEP) methods based on the conventional well logs data. Thus, after determination of dynamic Young’s modulus, some empirical correlations are proposed for estimation of static Young’s modulus. The results demonstrate that the Jambunathan equation is more appropriate than other empirical models. Finally, artificial intelligence (AI) techniques were run, and their results indicated that all techniques (BPNN (R=0.999), SVR (R=0.997) and GEP (R=0.996)) deliver highly accurate values of static Young’s modulus. Comparing these results shows that the BPNN technique is relatively more precise than other ones. Although, in this research, the GEP technique was not more accurate than BPNN and SVM techniques, it provides a new nonlinear equation that can be used for estimating Young’s modulus in other similar fields. As a new finding, it was found that a simultaneous combination of the Jambunathan equation and BPNN technique delivers highly accurate results. Hence, it can be applied to slim down the cost of exploratory operations for determination of the Young’s modulus of limestone rocks.

DT log is one of the most frequently used wireline logs to determine compression wave velocity. This log is commonly used to gain insight into the elastic and petrophysical parameters of reservoir rocks. Acquisition of DT log is، however، a very expensive and time consuming task. Thus prediction of this log by any means can be a great help by decreasing the amount of money that needs to be allocated for acquisition. Support vector machine (SVM) is one of the best artificial intelligence techniques proven to be a reliable method in the prediction of various real world problems. The aim of this paper is to use SVM to predict the DT log data of a well located in the southern oilfields of Iran. By comparing the results of SVM with those obtained by a Back Propagation Neural Network (BPNN) we were able to verify the accuracy of SVM in the prediction of P-wave velocity. Hence، this method is recommended as a cost effective tool in the prediction of P- wave velocity