Alltem Uxo Detection Sensitivity And Inversions For Target Parameters From Yuma Proving Ground Test Data

ALLTEM is a multi-axis electromagnetic induction system designed for unexploded ordnance (UXO) applications. It uses a continuous triangle-wave excitation and thus measures target step response rather than the more common impulse response. In May, 2006, we operated ALLTEM over the Army’s UXO Calibration Grid and Blind Test Grid at the Yuma Proving Ground (YPG), Arizona. The system multiplexed through all three orthogonal (Hx, Hy, and Hz axes) transmitting loops and recorded a total of 19 different transmitting (Tx) and receiving (Rx) loop combinations. This was accomplished while in continuous motion with a spatial data sampling interval, after waveform averaging, of 15 cm to 20 cm. ALLTEM records data at a constant 100 kilosamples/s rate at 24-bit precision rather than in a small number of time gates. Maps produced from differences between early- and late-time unfiltered signal amplitudes are of high quality, but exhibit ground response and system thermal drift effects. We find that by exploiting the high density of the time-series data to digitally filter the data and by moving the early- time pick to 275 μs, late enough that the step response of an analog low-pass filter has settled, instead of the 75 μs early time we had previously used, we can produce amplitude difference data and maps that are almost free of ground response and system drift effects while retaining good sensitivity to UXO. The improvement in the signal-to-noise ratio (SNR) greatly enhances the ability to detect small or deep targets and also produces more consistent inversion results from ALLTEM data. An inversion algorithm has been developed and applied to data from various sets of the available 19 Tx-Rx combinations over a number of targets. We present results that show that the inversion algorithm produces accurate parameters for some known targets in the Calibration Grid. This suggests that it is possible to obtain good multi-axis system target inversions from moving platform data even with some position “noise.”