During the deposition of fluvial sand body

various lithological assemblages can be formed due to different spatial distribution and thickness of sandstones and mudstones

and their corresponding seismic reflections have different waveform patterns.The technique of the neural network waveform classification can be used to analyze and summarize their seismic waveforms

and then predict regional distribution of sandstone sedimentary facies.Considering the low resolution of the seismic data and the thin thickness of sand layers

it is necessary to do a band spread processing with fidelity.However

this processing should be based on higher correlations between wells and seismic data and stable lateral amplitude trends.According to the sedimentary characteristics of the 3rd member of the Permian Xiashihezi Formation in Daniudi gas field in the Ordos Basin

we performed correlation and classification of seismic waveforms which reflect different lithological assemblages by using the neural network waveform classification technique

and predicted the distribution of several sandstone sedimentary models.Comparison with well data-based sandbody distribution map shows that they are consistent.The technique is quick and effective for sandbody prediction and has a better resolution for thin sand bodies.