5.1.6. IP Inversion¶
Here, demonstrate how to use dcsensitivity.exe and sens2weights.exe to compute sensitivity weights for the IP inversion. We then use the code ipoctree_inv.exe to recover a chargeability model. Because this is a simple example with no noise, we assigned uncertainties of 0.0005 to all IP data. In practice, data are noisy and choosing appropriate uncertainties is very important for successful inversion.
Note
Depending on the application, sensitivity and/or interface weighting may or may not improve the final model. We apply the weights to promote familiarity with this code. For this example, the data are very-well constrained by the data and significant weighting may not even be necessary.
5.1.6.1. Sensitivity Weights¶
Here, the code dcsensitivity.exe and the input file ipsens.inp (see format) are used to approximate the sensitivities for the IP problem. Then the code sens2weights.exe and the input file sens2weights.inp create a sensitivity weights file. This counteracts the inversion’s natural tendancy to incorrectly place anomalous structures near the electrodes. Files relevant to this part of the example are in the sub-folder ip_sensitivities . Before running this example, you may want to do the following:
To compute the sensitivities, the following input file was used:
To generate the sensitivity weights file, the following input file was used:
The final sensitivity weights for the IP inversion is shown below.
Important
Although we demonstrate how to compute sensitivity weights here, we will not include them in the IP inversion.
5.1.6.2. IP Inversion¶
Here we use the code ipoctree_inv.exe to recover a conductivity model. Before running this example, you may want to do the following:
Files relevant to this part of the example are in the sub-folder ip_inv. To invert the synthetic data, the input file below (ip_inv.inp) was used. For formatting, see format :
The true model (left) and the final recovered model (right) are shown below.
