神经网络自变量输入模式的视电阻率求解算法

根据瞬变电磁场理论公式中的响应和自变量之间的关系特点，提出用非线性方程模式的 BP 神经网络求解电阻率。通过构造单输入单输出网络结构，建立以不同时间点上的电流归一化的感应电压值为输入、视电阻率值为输出的神经网络，来拟合瞬变电磁场的二次涡流曲线。利用数值方法计算出的数据验证该训练网络的精确性，比较了不同算法对训练精度和收敛速度产生的影响。以重庆大学某处的防空洞探测实验为例验证了该算法的有效性，该算法避开具体的复杂电磁场计算或数值反问题计算，从而实现电阻率快速计算，为快速成像准备必要条件。

Solution algorithm of apparent resistivity based on independent variable input mode of neural network

According to the relationship between the response and the independent variables in transient electromagnetic field theory, a BP neural network with nonlinear equation model was proposed to solve the resistivity. By constructing a single-input-single-output network structure, a neural network with current normalized induced voltage at different time points as input and the apparent resistivity as the output was set up to simulate the secondary eddy current curve of the transient electromagnetic field. The accuracy of the trained network was verified by the data calculated by numerical computation, and the training accuracies and the convergence speeds of different algorithms were compared. The effectiveness of the proposed algorithm was verified by the experiments in an air-raid shelter in Chongqing University. The presented solution algorithm avoids calculation of complex electromagnetic field or numerical inverse problem, and realizes fast calculation of resistivity.