Highly accurate seismic inversion results help refine geomechanical modeling. In this paper, a post-stack seismic sparse dictionary learning inversion method based on logging data is proposed. First, feature functions are extracted from the logging data. Then, a dictionary is learned adaptively from known observations. This dictionary is composed of a series of feature functions, using which the parametric model can be effectively characterized. This method effectively avoids the problem of single mathematical model assumptions. Finally, the post-stack wave impedance inversion data are solved, and the separation of wave impedance data is performed using the correlation between the velocity data and the density data. This method can effectively improve the resolution of seismic inversion results by extracting a priori information from logging data. It is found that the root-mean-square error of the sparse dictionary learning method is reduced by 9.075% compared to the Tikhonov method.
Due to the lack of drilled well data in the undeveloped area of the field, the only available information is seismic data. The seismic data effectively reflects the rich stratigraphic characteristics such as lithology, tectonics, and physical properties in the study area. However, the seismic data need to be analyzed and converted through a series of processing to obtain velocity data and density data of the study area which can be used as input data for geomechanical models. Currently, there are several methods of using seismic data to obtain the basic input data for geomechanical models.
In conventional studies, the Dix formula is generally used to calculate the layer velocity as a basic geomechanical data (Dix, 1955; Gholami et al, 2019). The method is mainly based on the stacked velocity obtained from the raw seismic data. After dip and phase correction, the stacked velocity is converted to layer velocity. This method is simple and applicable. However, the accuracy depends on the stratigraphic position and the accuracy of the stacked velocity. In particular, the accuracy of this method is not high when the dip angle of the stratum and the lateral variation of the velocity are large.
