摘要
重力勘探资料是不同深度地质异常体的综合反映,位场分离工作向来都是重力勘探的重要环节。本文首先通过构建BP神经网络模型和趋势分析模型来拟合低频区域重力场,进而分离出局部异常场信息;然后通过理论模型的分离结果对比分析,对于中浅部的局部异常体信息,BP神经网络模型分离结果精度高,其短波长信息源更精确。最后,将BP神经网络模型运用到南非Witwatersrand Basin某实测航空重力测网数据分析中,其分离结果的Theta map成像结果较好地吻合了实际地质资料,验证了BP神经网络模型在实际重力勘探解译中具有一定的适用性。
Gravity exploration data is a comprehensive reflection of geological anomalies in different depths,and the separation of potential fields has always been an important part of gravity exploration.This paper built a BP neural network model and a trend analysis model to fit the low-frequency region gravity field,and then separated the local abnormal field information.Through the comparative analysis of separation results of the theoretical model,for the local abnormal body information in the middle and shallow parts,the separation results of BP neural network model were significantly more precise than the trend analysis model,and its short-wavelength information source is more accurate.Finally,the BP neural network model was applied to the data analysis of the airborne gravity survey network measured in the Witwatersrand Basin in South Africa.The theta map imaging result of the separation result is in good agreement with the actual geological data,which shows that the BP neural network model has certain applicability in interpretation of actual gravity prospecting.
作者
杨磊
Yang Lei(Geological Subgrade Design and Research Department, China Railway Siyuan Survey and Design Group Co.,Ltd.,Wuhan Hubei 430063, China)
出处
《工程地球物理学报》
2021年第1期90-97,共8页
Chinese Journal of Engineering Geophysics
