Abstarct: To model a reservoir, it is necessary to accurately evaluate the properties of rocks and visualize their heterogeneity. Due to high drilling costs and the lack of sufficient large-scale wells or data related to wells and cores, 3D seismic data play a very important role in detecting reservoir changes. Any change in lithology, porosity, and fluid content causes changes in amplitude, frequency, lateral cohesion, and other seismic markers. If these changes are detectable in seismic parameters, they will help to understand subsurface geology. The purpose of seismic facies analysis is to interpret changes in seismic parameters. Seismic facies can be defined as seismic rejection groups. Many methods have been developed to identify and detect patterns for seismic facies. An unsupervised pattern is used when geological information is not available. The presence of noise is an integral part of seismic data that affects the classification of facies and seismic interpretation, and its presence causes misinterpretation. The research approach for this problem is multi-resolution seismic facies analysis. This means that before the main seismic data or markers enter the facies analysis process, we first decompose the seismic data into its constituent components using one of the signal decomposition methods and then participate in a number of components in signal reconstruction and some components. We leave out the ones that express the nature of noise. Then a reconstructed version of the signal or the resulting markers is entered as input to the facies analysis algorithm. In this research, the intention is to decompose seismic data into their constituent components using variational mode decomposition (VMD) and to reconstruct the facies analysis process on the data and apply the resulting markers. Facies analysis in this method will be performed without supervision and by K-means method. The performance of artificial and real data methods is tested and it is shown that more accurate facies analysis can be performed.