Category : New publication
Congratulations to Mohammad Amin Partovi, Msc student of petroleum engineering, for his recently published work about geological boundary detection from well-logs using pattern recognition techniques in Petroleum Science & Engineering.
In this study, we implemented an automatic well-to-well correlation approach based on pattern extraction from well-logs. The well-log patterns were recognized by calculating several statistical and fractal parameters. As a fractal parameter, we selected the wavelet standard deviation exponent, calculated by the discrete wavelet transform. Furthermore, to select the proper wavelet function, the energy to Shannon entropy ratio criterion was implemented. The statistical pattern recognition parameters of this study include average value, maximum to minimum ratio, coefficient of variation, and the trend angle of well-log data (i.e., Gamma ray log) in a window around the geological boundary. Moreover, the analysis of variance (ANOVA) tool and the Tukey multiple comparison method were implemented to evaluate the effectiveness of each parameter (i.e., fractal or statistical parameters) during the determination of boundary depths. The gamma ray logs from three wells of an oil field were used as a dataset for the evaluation of our algorithm. The outputs of our methodology were also compared with the new detrended fluctuation analysis (DFA) method, which show promising outcomes over it. The results show that the average value of the signal, among the analyzed parameters, is the most effective parameter; however, implementing the combination of all fractal and statistical parameters can improve the accuracy of the geological boundary detection.