The research described in this report was conducted in fulfillment of Project MR-2728, “Enhanced EMI Models and Systems for Underwater UXO Detection and Discrimination”, submitted to the Strategic Environmental Research and Development Program (SERDP) in response to Statement of Need # MRSON-17-01 Detection, Classification and Remediation of Military Munitions Underwater (UW)”.

The main objectives of this report are threefold:

  1. To model, analyze, and mitigate conducting environment electromagnetic responses that can affect the performance of advanced electromagnetic induction (EMI) sensor arrays in the marine environment.
  2. To develop enhanced physically complete EMI models that will accurately account for EMI responses from marine environments, for UW targets detection and classification.
  3. To demonstrate applicability of the enhanced model for UW targets classification by processing and analyzing UW data sets.

Technical Approach

To achieve these objectives, first, the numerical methods, such as the unconditionally stable Crank-Nicolson finite different time domain (FDTD), the method of auxiliary sources (MAS), and a semi-analytical cylindrical plane wave expansion model were adapted to UW EMI sensing problem; then the studies were done for the primary, secondary, and reflected electromagnetic fields in conducting and permeable multilayer structures; third, comparisons between modeled and experimental data sets were given for an advanced transient EMI sensor placed in marine environment; fourth, the enhanced UW transient EMI models (such as updated dipole model, transient image technique) were developed and illustrated for accurate modeling of UW geophysical data sets; and finally an EMI data set that was collected via an advanced EMI array in the marine environment, was processed and targets classification feature parameters were extracted and analyzed.


The analytical, numerical, and limited-experimental studies have shown: 1) marine environments distort both the primary and secondary magnetic fields at early times/high frequencies, that signal distortion is a function of separation distances between the target and the Tx coil and between the target and observation points. As the distance between target and sensor increases, distortion of the target’s EMI signals at later times is observed; 2) water-air and water-sediment boundaries produce non-negligible EMI responses, which vary with respect to the transmitter loop size, sediment conductivity and magnetic susceptibility, as well as the distances between the loop and water-air and water-sediment boundaries; 3) direct signal from the Tx current to Rx coils (i.e., salt water effects) influence a target’s EMI responses disproportionately. The effects depend on the target’s properties and its distance from the Tx and Rx; the comparisons between reflected (secondary) signals and the response from a conducting, permeable and non-permeable sphere/105mm projectile illustrate that the background signals (total response from layer boundaries and direct signals from Tx) are higher than the response from the sphere/projectile at early times; 4) the surrounding marine environment produces time domain EMI signals (Lenz’ Law), which can outweigh the target’s responses for large separations between the target and Tx/Rx. Among these effects, the distortions of the primary magnetic fields at receiver (direct coupling from Tx to Rx) have a dominant influence on the UW target’s EMI responses. To mitigate these effects, enhanced EMI models were developed. The enhanced models account for UW environmental effects accurately and extract target specific classification feature parameters from actual UW EMI data sets. 


The enhanced UW EMI data inversion and classification algorithms were applied to single-pass ultra-transient electromagnetic array data sets collected at Sequim Bay, WA UW calibration grid. First, background levels were removed from the survey line data for each time channel using detrend algorithm with 10 m window. Once EMI data were leveled, then two approaches were deployed to pick anomalies for further interrogations: (1) The traditional method that utilizes signal amplitudes on a 2D map and identifies peaks of signals above a prescribed threshold level; and (2) A semi-supervised Gaussian clustering process which clusters the inverted extrinsic (source locations) parameters into a 3D space and identifies targets using cluster centers. In the latter approach, the combined Orthogonal Normalized Volume Magnetic Source – Differential Evolution algorithms were applied to each dynamic data point and the intrinsic and extrinsic parameters of the anomalies extracted using a multiple source inversion approach. The calibration data analyzes have showed that the enhanced EMI models can locate and identified all TOI.