User's Manual
!"#
$%+&"
thicknesses, and different types of gradation at the layer boundaries. All of these properties
affect the characteristics of the reflected radar waves and a change in any property can change
the reflected radar waves. A traditional approach to analyzing radar data is to use a physics-
based approach that specifically accounts for each of the properties listed above; but this is
difficult for such a complex problem. Neural networks are an alternative approach that is well-
suited for problems involving complex pattern recognition. The RABIT system uses a neural
network that has been trained using data from numerous field outings where physical ballast
samples were collected and analyzed. The neural network examines each GPR waveform and
estimates the ballast fouling (i.e. Selig fouling index [1]) and moisture for the depth interval 0-16
inches (0 to 40.6 cm). The details of the neural network training and analysis can be found in
[2].
5 Hardware Assembly
During normal storage and transportation, the RABIT is packed in its shipping case. To
unpack from the shipping case and assemble the unit, refer to Figures 2, 3, and the steps
below.
1. Open the shipping container and remove the ‘T’ handle and the three outriggers (Figure
2, right).
2. Remove the main scanner frame including the white antennas (Figure 2, left).
3. Place the frame assembly with the sensor units on the ground so that the white antenna
boxes are on the ground. Attach the outriggers as shown in Figure 3. Make sure that all
push pins are fully inserted.
4. Attached the push rod to the attachment point above either wheel. It doesn’t matter
which attachment point (which wheel) is used.
5. Operate the unit as described in the Operation section. When finished, disassemble the
frame in the reverse order and return the components to the shipping case.
Figure 2. The RABIT and its shipping container from the shipping container.