GNGTS 2017 - 36° Convegno Nazionale

642 GNGTS 2017 S essione 3.2 is projected into the small dimensional subspace (the coarse grid) in which the problem is solved with a direct method, while the component of the solution in the remaining subspace is computed by an iterative algorithm. Experimental set-up. The experiment was conducted in a Plexiglas cylinder having a height of 250 mm and a diameter of 500 mm. Fig. 1 depicts the experimental setup, showing the data locations and the position of the reference electrode in all data sets. A total of 49 holes were drilled through a Plexiglas plate every 5 cm in an 7×7 grid and used as a template for electrode placement and manual insertion at five elevations. The tank was filled with silica sand to depth of 20 cm by taking care to avoid air entrapment. A metallic body consisting of a rectangular piece of iron with a thickness of 2.5 cm and a height of 11 cm was centered at the phreatic surface. The phreatic surface was maintained constant by adding small amounts of water to the base of the tank through a vertical plastic tube installed prior to filling with sand. Experiment began by first taking measurements at the surface of the tank at the positions indicated in Fig. 1b and then collecting data along the cross section shown in Fig. 1a to further explore the nature of the SP signals, once significant surface anomalies were observed to develop. These data were manually collected throughout the tank at 4-cm-depth intervals from 0- to 20-cm depth. For SP data acquisition, Ag/AgCl nonpolarizing electrodes (pellet electrodes with a diameter of 2 mm, a length of 4 mm and a exposed wire of 70 mm) and a calibrated voltmeter (ABEM Terrameter SAS300 resistivity meter with a sensitivity of 1 μV and an internal impedance of 10 MΩ) were used. Bulk electrical-conductivity distribution in the tank was determined by performing a 2D resistivity tomography using Syscal-Pro resistivimeter (IRIS Instruments) with 10 channels and a dipole-dipole array. Resistivity data were collected utilizing 17 gold electrodes along a profile and 2.5-cm spacing between electrodes and were inverted with the VEMI algorithm built within the EIDORS environment (De Donno, 2013). Results and discussion. Self- potential anomalies due to the corrosion of the iron bar were measured for several weeks at the top surface of the tank (Fig. 2). On the 8th week of experiment, the SP depth data collection was conducted throughout the tank showing a dipolar self-potential distribution, with a positive anomaly located in the bottom part of the iron bar and a negative anomaly located in the vicinity of the top part of the metallic object. Fig. 3 illustrates the results of 2D inversion of surface electrical-resistivity measurements. Resistivity and depth SP data were used as input for the inversion to localize the causative source body. In the first step, the kernel computation accounted for the electrical resistivity distribution and for the insulating boundary conditions applied to the system and governed Fig. 3 - 2D resistivity tomography. The resistivity ranges from 200 ohm-m in the saturated portion of the tank (and in the capillary fringe) to nearly 3500 ohm-m close to the top surface of the tank.