Because of the complicated pore structure and complex relation of oil and water, especially owing to the existence of high-viscosity oil and bitumen in the formation it makes the resistivity difference between oil and water zones very low. So it is very difficulty to identify the oil and water with logging data. By applying the capillary pressure curves, lithological- electrical experiments and relative permeability curves the Silurian reservoir classification was studied in combination with the reservoir fractal characteristic and a classification model was established by using non-linear method to identify the reservoir types. On the basis of classifying reservoir types a improved normal distribution method was presented to distinguish oil and water zones. Then with core data the different parameter interpretation models concerning to varying reservoir types were founded. By used in practice it was proved to be effective to improve the veracity of log interpretation for bituminous reservoir in Tarim basin.