Objective

The Army demilitarizes non-usable ammunition in the rotary kiln incinerator (RKI) and tries to recycle the scrap through commercial dealers. The scrap must be safe to leave government custody, Success is meeting or exceeding the human 200% ability. Additional insight will be obtained since the x-rays can “see through” the ammo casing and projectile for energetics. Obtaining approval from the Department of Defense Explosive Safety Board (DDESB) will allow an alternate means of inspection, using advanced technology. Automation of the inspection process is another objective of the effort, and the better judgment produced by the AI & ML will increase the sorting effectiveness. Current and emerging requirements (such as the need for “venting” cavities and voids that can’t be “seen”) increase benefits of x-ray penetration and imaging. The following are the topic objectives: detect and discriminate for energetics remaining in ammo scrap and sort as MDAS or MPPEH; create a system of systems that can judge energetic levels as good as or better than human inspection for scrap from the demilitarization (demil) furnace.

Description

The U.S. Army has a need for reliable, accurate, and repeatable detection of explosive hazard on metallic ammunition scrap. Current inspection methodologies utilize two independent inspectors, who conduct a ‘200%’ visual inspection which poses costs, safety hazards, accuracy levels, and limits of vision. This also requires additional time with inspection results that are prone to human error. The program builds on the advances in x-ray technology, digital imaging, and advanced algorithms. The project will use artificial intelligence (AI), machine learning (ML), deep learning, and neural networks to interpret metallic scrap inspection images, that are produced by x-rays, for traces of explosive hazards. Success will significantly reduce operator workload and allow a more efficient and effective means of determining pure metal scrap. The results will be integrated into automated sorting for MDAS (Material Documented as Safe) from MPPEH (Material Potentially Possessing Explosive Hazard) Success, and advancement, will be the ability to inspect as good or better than humans reliably and repeatedly, and also “see inside” of questionable surfaces. If this is successful, it could mean a new alternative for inspections, less manpower to be expected, and better documentation for scrap output.

Phase I

Phase I will consist of demonstrating the feasibility of an advanced AI algorithm using (labeled or unlabeled) x-ray images taken from metallic scrap ammunition. This will take algorithms from training sets to create convolutional neural networks, regression models, and unsupervised learning. Phase I should increase effectiveness of the ASI project.

Phase II

Phase II will further mature the AI/ML technology to meet all requirements in the system, with a final demonstration in a relevant environment.

Phase III

Success of Phase II would allow transition to more ammo depots and larger sizes.

For more information, and to submit your full proposal package, visit the DSIP Portal.