Optimization determine of petrophysical facies class using Fast independent component analysis and discrete cosine transform based k-nearest neighbors in the Marun oil field, Asmari reservoir

This paper aims to optimally determine petrophysical facies according to well log data. Using the automatic classification method of K-NN (K-Nearest Neighbours), petrophysical facies can be determined even though not optimally. For optimal determination of facies, the K-NN method is combined with FastICA (Fast Independent Component Analysis) and DCT (Discrete Cosine Transform) methods. This increases the success rate of the K-NN method. It also brings about optimal determination of petrophysical facies after which modelling and description of hydrocarbon reservoirs can be done. The research is performed in two different ways: In the first approach, the FastICA method is applied to data and then classified by the K-NN method. In the second approach, FastICA and DCT methods are applied to data and then classified by the K-NN method. Finally, the success rate of classification by the K-NN method is evaluated in both approaches to optimally determine petrophysical facies. Such evaluations indicate that application of the second method to data significantly enhances the success rate of the classification by the K-NN method, thereby leading to optimal determination of petrophysical facies, which is the very aim of this study. The utilized data including sonic log (DT), gamma rays (SGR), density (FDC or RHOB), neutron porosity (CNL or NPHI), and deep induction logs (ILD), belongs to the Marun oil field in southern Iran.