The objective of this project was to demonstrate a system capable of providing the precise relative position of an electromagnetic induction (EMI) sensor while recording data over a buried target. EMI data acquired as a function of position can be used as input into sophisticated target characterization algorithms that require spatial sampling. The basic approach is to use the active transmitter of the EMI sensor as a beacon that can be located and oriented in space relative to a set of fixed reference coils that detect the sensor transmitter field.
In the course of this work, two beacon location algorithms were investigated based on this principle. The first algorithm required beacon signals recorded by two sets of orthogonal coils to compute the (x, y, z) coordinates and the orientation of the active sensor. This approach required a nonlinear, iterative least-squares algorithm, which was relatively computer intensive. The prototype positioning system had two sets of three orthogonal coils housed inside each end of the “beam.” The separation between the two sets of coils was about 1.8 meters. In this system, the beam needed to be elevated above the plane on which the sensor moved to allow flux linkage through the vertical coils (i.e., the coils whose axes point horizontally).
A second, simpler algorithm was more recently developed, which required only three horizontal coils (axes vertical). The latter configuration assumed that the coils and the sensor were in the same horizontal plane, i.e., it solved for the (x, y) coordinates of the sensor position and assumes z = 0. The significant advantage of this algorithm was that a closed-form, analytical solution existed for computing (x, y), which eliminated the need for an iterative algorithm. The orientation of the sensor was not computed using this approach; however, its accuracy was independent of the sensor orientation. A third (center) coil was used to normalize the end coil measurements. It was shown that the sensor orientation dependence disappears as a result of this normalization. The only assumption was that the coils and sensor center were co-planar (although the sensor transmitting coil need not be horizontal). Simulations have shown, however, that the accuracy of the (x, y) calculation was fairly insensitive to small (e.g., 5 or 10 cm) deviations of the sensor from the plane of the three reference coils.
The positioning algorithms required accurate measurements of the amplitude of the sensor signal by the reference coils. Accurate estimations of this amplitude were more difficult to achieve using the EM-61HH due to the greater complexity and short duration of the time-domain pulse compared to that of the continuous-wave GEM-3 signal. Part of the challenge in estimating the amplitude of the EM-61HH waveform was in the selection of a time window over which the amplitude integration was to be performed. The optimal duration and location of this window were more difficult to determine as the signal becomes noisier at greater sensor ranges. An advantage of the GEM-3 instrument was its excellent signal-to-noise performance, which arose from the narrowband quality of the signals. The sine and cosine multiply and integrate process performed by the GEM-3 over many cycles greatly reduced the effects of broadband noise. Broadband time-domain systems are, however, inherently more sensitive to broadband noise, although high signal levels can compensate to some extent for this sensitivity.
Results were presented using the new analytical algorithm employing data recorded with a prototype of the three-coil system. Such a system has been constructed and tested using both EM-61HH data and GEM-3 data. Good results were obtained with both sensors, although somewhat superior positional accuracy was achieved with the GEM-3 system, which is likely due to the narrowband character of its signals and the ease of the processing of such signals, as opposed to the more complex waveforms generated by time-domain systems.
The technology could enhance target classification by providing precise multispatial sample position information over a target for use by advanced data analysis algorithms, using less expensive and less cumbersome hardware than other systems on the market.