ISE Project Helps DOD Manage Parts Supply Chain

If a part for an old refrigerator in your garage is no longer made, it's easy to replace the entire refrigerator with a new one. If the same thing happens with a part for a military submarine or helicopter, the outcome can be much more dire and expensive.

Researchers in the Department of Industrial Systems and Engineering at the University of Tennessee are collaborating on a project for the U.S. Department of Defense to develop a knowledge-driven and analytics-driven conversational AI agent that can assist analysts with managing parts obsolescence. Parts obsolescence occurs when parts become unavailable due to causes such as suppliers going out of business, suppliers being acquired by a larger company, and changes in environmental and safety regulations.

The hope is that the AI agent will be able to reduce the time and effort required to identify supply chain vulnerabilities and find cost savings opportunities that a human may miss because of the vast number of parts managed by the DOD at the enterprise level.

"These are systems where it's not just a preference to stay up and running; it's a non-negotiable for them to be up and running," said Chad Uhles, a fourth-year Ph.D. student working on the project under ISE Professor and Principal Investigator Jim Ostrowski. "Whenever you have those significant amounts of system downtime, you're going to incur risk well beyond outside of the system. If you look at it from another game theory idea, what can an adversary do in the time that the system is down? How can they take advantage of it?"

Creating an Efficient System

Uhles, who is doing his thesis at UT on obsolescence management, says organizations have three main ways to combat the issue: they can stock up on supply, try to find an alternate supplier, or redesign the system to use an entirely different part.

"The military is increasingly running into this issue because some systems were designed in the '50s, '60s and '70s and they didn't think they were going to make it to year 2025," Uhles said. "If something goes out on these machines, the company that made the part may not be in business anymore. So where are we going to be getting these parts from? Or how are we going to be able to identify substitutes?"

The AI agent will be designed to improve readiness by preventing delays in repair and maintenance caused by outdated or unavailable parts. This process will happen through integration into a parts management system, providing the system with advanced decision-support capability.

Given the age of some of the equipment used in the military, a part could go obsolete in five months or in five years.

"Because there's so much uncertainty in that window, you're trying to balance the amount of inventory that you have," Uhles said. "You don't want to have too much inventory on hand, while also making sure you have enough to get to whatever goal that you have set for system support. So that's really the balance that we're striking right now."

Cost-Saving Strategies

The team involved in the DOD project is in the early phase of generating ideas to determine optimal spare part inventory levels that will mitigate obsolescence in the most cost-effective way.

Uhles is working on translating the project's stated goals into math and then translating the math into code through Gurobi Optimizer, a high-performance solver for mathematical programming.

"It can handle hundreds of thousands and even millions of variables and be able to spit out solutions," Uhles said. "From there, it's up for our interpretation to determine if the results look right. Is it possible that the model may be doing something that it's not really supposed to do? Then, you start iterating over that process of how we can improve the solutions that we're getting."

As they receive more data from the military, Uhles and the research group can develop more precise algorithms and optimization models that will help streamline the DOD's supply chain and greatly impact the bottom line.

"These are multimillion dollar systems that we're talking about, sometimes even the tens of hundreds of millions of dollars," Uhles said. "There are significant cost savings that can be made by being able to manage our spare parts inventory and creating more efficient and more robust ways that we can communicate between military branches."

Provided by University of Tennessee at Knoxville