Objective In order to identify the key geological structures in the metallogenic process of coal measures metal deposits, the Linxing block of Ordos Basin was taken as the research object. Based on the previous research results, geological data of the block, seismic and logging data, a geological model including tectonic volcanic activities and other events was constructed to analyze the key geological factors of deposit formation.

Methods Firstly, based on the post-stack seismic data, the diffraction wave separation and imaging are realized by using the plane wave destruction filter technology, and the small-scale key geological information is extracted. Secondly, the attribute optimization is carried out by using the knowledge map combined with the actual situation of the study area, and the dominant attributes reflecting the structural characteristics such as variance, instantaneous frequency and root mean square amplitude are obtained. Taking the three-dimensional seismic attribute volume of the study area as the training sample, the mapping label is established based on the knowledge graph, and the U-Net multi-attribute fusion model is proposed by introducing the convolution block attention mechanism. The model uses three-channel seismic attributes as input and underground key geological structures ( such as faults or volcanic channels ) as output, and then constructs an intelligent identification method for key geological structures.

Result and Conclusion The results show that the U-Net network based on the knowledge map has high efficiency and accuracy in the identification of key geological structures of coal-bearing metal deposits. It can intuitively depict the three-dimensional distribution characteristics of key geological structures, reduce the artificial uncertainty of fault interpretation, and provide effective technical support for the study of coal-bearing metal deposits.