Usually, the logging recognition method for sylvite layer is mainly by means of radioactive logging whose working effect is good without the influence of other potassium minerals. According to the geological data, Guang'an area is defined as a polyhalite deposit zone, which has multiple potash layers comprising mainly polyhalite. In addition, there are some other potassium rocks in this area, such as mudstone, argillaceous dolomite, magnesite mudstone and volcanic tuff, which all have high-level radioactivity that causes difficulties in dividing the sylvite layer with radioactive logging. On the basis of the differences between various logging curves of potassium rocks, the authors identified the sylvite layer by such means as logging curve comprehensive analysis, logging curve overlapping, cross-plot analysis and NGS discriminant model and divided polyhalite layers of somewells in different structures of Guang'an area in central Sichuan. A comparison with logging data shows that these methods have good effects and are feasible. The authors made a statistic analysis of the logging response and summarized the distribution layers and the effective thickness of polyhalite, and the results obtained enable us to learn more about the distribution of polyhalite in Guang'an area and expand the use of geophysical logging data in potash exploration.