基于BP神经网络的Cole-Cole模型参数预测

Cole-Cole Model Parameter Prediction Based on BP Neural Network

由于Cole-Cole模型的频谱参数可以用于区分引起激电异常的极化体和寻找深部矿(化)体,国内外学者采用复电阻率数据对频谱参数的确定做了大量的工作,并取得了很大的成果。而传统正演方法存在计算量大,面对复杂地形正演困难等,同时为了更好的准确识别矿体,减少误判。本文根据BP神经网络训练可直接通过输出预测值来代替传统正演算法,这可大大减少传统正演的时间,提高计算效率。预测结果验证了BP神经网络学习在理论上Cole-Cole模型参数反演是可行的、有效的,可为下一步的地球物理勘探工作提供更为丰富的参考价值及指导性意义。

Since the spectral parameters of the Cole-Cole model can be used to distinguish polarizers that cause excitation anomalies and find deep ore bodies, scholars at home and abroad have done a lot of work on the determination of spectral parameters by using complex resistivity data, and have achieved great results. However, the traditional forward method has a large amount of calculation, and it is difficult to face complex terrain, and at the same time, in order to better accurately identify the ore body and reduce false judgment. According to BP neural network training, this paper can directly replace the traditional forward algorithm by outputting the predicted value, which can greatly reduce the time of traditional forward evolution and improve the computational efficiency. The prediction results verify that BP neural network learning is feasible and effective in theory for Cole-Cole model parameter inversion, which can provide richer reference value and guiding significance for the next geophysical exploration work.