One of the major overarching goals of this project is to make headway toward determining how to best establish the necessary confidence in system performance in order to convince regulators of a method’s efficacy. The diversity and variability in the underwater unexploded ordnance (UXO) remediation problem is massive, and thus the project team focuses attention first on the UXO-side of the problem. The objectives are distilled down to three main topics: 1) arrive at a set of well-characterized UXO (surrogate and inert), 2) use validated modeling techniques to quantitively analyze system performance as a function of variability in UXO characteristics, and 3) using model and real data, assess how precisely the characteristics of surrogate UXO must match those of inert UXO.
The approach is comprised of four major tasks. First, the project team will work on procuring surrogate and high fidelity inert UXO of interest, specifically looking to establish open dialogue between shell manufacturers and munitions experts to better understand the implications that manufacturing procedures have on the acoustic properties of the UXO. Next, the project team will investigate methods for characterizing UXO using various nondestructive testing techniques, as well as free field acoustic measurements. The characterization will be performed on both the high fidelity inert UXO, as well as the surrogates, paying specific attention to how well the physical and acoustic properties of the surrogate match the inert UXO. Computer-aided design models will be built of the UXO, sufficient for use in finite element (FE) models. In addition to building high fidelity FE models, multiple versions will be built that include variations to the UXO properties (e.g. shell thickness, diameter, material of shell and filler, internal components). The free field FE models will be used with the Target-in-the-Environment Response model to simulate “through-the-sensor” time series acoustic data that mimics what would be collected with the Multi-Sensor Towbody (MuST). In addition to variations in the UXO characteristics, these synthetic data sets will also explore various environmental conditions. Convolutional Neural Network classification strategies currently in use with the MuST system will be used to quantify the effects to performance as a function of variations in the UXO properties. Finally, field data collected during recent and future MuST deployments will be analyzed along-side the synthetic data, thereby allowing for a quantitative analysis into how closely the surrogates must match the corresponding higher fidelity inert UXO, again using system performance as the metric.
Ultimately, a set of well-characterized surrogate UXO will help developers and regulators establish expected performance bounds during demonstrations and live remediation exercises, specifically opening the door for performance assessment between different UXO remediation systems. An important aspect of this will be quantitatively assessing how precisely the surrogate UXO need to match the real UXO, using system performance as a metric. Understanding how performance is linked to the physical characteristics of the UXO will help SERDP and ESTCP make an informed decision regarding the level of future investment in surrogate design and manufacturing. Finally, the synthetic data sets created under this project will not only span the physical properties of the surrogates, but also variations due to different underwater environments and geometries. These data will be made available to the SERDP and ESTCP community and are sufficient for training and testing prospective platforms.