APPLICATION OF NEUTRAL NETWORK TO PREDICTING DEEP-SEATED VOLCANIC ROCKS IN THE SONGLIAO BASIN

Forward modeling of the volcanic cores taken from the deep Songliao Basin demonstrates that the gravity and magnetic anomalies of the rocks, which usually are not high, are a kind of subordinate anomalies superimposed on a strong background. In this paper, the integral - iterative continuation flat change curve was selected to enhance the signal of magnetic anomaly of the volcanic rocks, in order to balance the magnetic anomaly so as to eliminate the depth influence. Through the magnetic anomaly processing and interpretation, the information of the magnetic anomaly of the deep-seated volcanic rocks were extracted. Then the methods of oblique derivative, Euler deconvolution and others were adopted to delineate the distribution pattern of the deep volcanic rocks. Results suggest that apparent density,apparent susceptibility and the correlation coefficient of the above two are the best combination of parameters. The neural network fuzzy recognition is an effective method in recognition of volcanic lithology. A well constrained discrimination network was established for identification of the lithology of volcanic rocks. The method could be used as an effective reference for prediction of deep-seated volcanic rocks in other areas