基于BP神经网络的时域激电谱Cole-Cole模型参数反演及应用
|
杨海明, 姚卫星, 唐塑, 潘展超, 关力伟
|
Parameter inversion and application of the Cole-Cole model for time-domain induced polarization spectra based on the backpropagation neural network
|
YANG Hai-Ming, YAO Wei-Xing, TANG Su, PAN Zhan-Chao, GUAN Li-Wei
|
|
表1 BP神经网络训练频谱参数的相关系数R及均方误差Mse(8 000样本集)
|
Table 1 Correlation coefficient and mean square error of spectral parameters in BP neural network training (8,000 sample set)
|
|
参数 | c | m | τ | 模型结构 | R | Mse | R | Mse | R | Mse | 3-[3]-3 | 0.686 672 | 0.164 371 | 0.659 548 | 0.178 509 | 0.616 054 | 0.226 556 | 3-[5]-3 | 0.703 721 | 0.158 974 | 0.699 014 | 0.177 708 | 0.680 032 | 0.190 873 | 5-[5]-3 | 0.825 955 | 0.105 401 | 0.833 382 | 0.090 517 | 0.726 354 | 0.209 827 | 8-[10]-3 | 0.883 236 | 0.124 673 | 0.869 024 | 0.120 921 | 0.853 044 | 0.152 909 | 8-[15]-3 | 0.833 561 | 0.165 781 | 0.812 135 | 0.143 291 | 0.752 719 | 0.190 345 | 8-[3 3]-3 | 0.826 257 | 0.095 627 | 0.840 143 | 0.100 481 | 0.804 028 | 0.187 341 | 8-[5 5]-3 | 0.945 304 | 0.038 803 | 0.953 647 | 0.046 215 | 0.873 556 | 0.180 023 | 8-[10 10]-3 | 0.956 072 | 0.035 238 | 0.965 203 | 0.026 610 | 0.924 37 | 0.119 473 | 8-[15 15]-3 | 0.943 289 | 0.424 467 | 0.949 861 | 0.058 321 | 0.879 808 | 0.130 944 | 8-[3 3 3]-3 | 0.955 658 | 0.043 565 | 0.980 468 | 0.028 764 | 0.915 581 | 0.125 348 | 8-[5 5 5]-3 | 0.975 376 | 0.032 639 | 0.982 149 | 0.027 649 | 0.973 748 | 0.071 334 | 8-[10 10 10]-3 | 0.992 135 | 0.018 607 | 0.997 192 | 0.011 095 | 0.996 653 | 0.025 730 | 8-[15 15 15]-3 | 0.979 347 | 0.033 456 | 0.969 457 | 0.037 855 | 0.968 965 | 0.085 321 |
|
|
|