You are here

Using Artificial Intelligence to Optimize Missile Sustainment Trade-offs

Description:

TECHNOLOGY AREA(S): Weapons

OBJECTIVE: Develop methods to use artificial intelligence (AI), machine learning, and real-time computational intelligence to optimize Army logistics and sustainment simulations and predictions for both legacy and future Army missile systems.

DESCRIPTION: The CCDC AvMC Logistics Engineering Lab (LogLab) developed a sustainment simulation capability for Army aviation using a government-owned software tool called System of Systems Analysis Tool (SoSAT). Multiple PM's use this capability to conduct analysis and provide input for major acquisition documents. The LogLab is looking to upgrade the simulation capabilities of the software tool using artificial intelligence and machine learning to optimize logistics outcomes for CCDC AvMC customers like Hypersonics. Artificial intelligence would determine strategies of sparing, costs, supply chain locations, maintenance staffing, maintenance levels, scheduled maintenance times, to best measure and optimize sustainment options and logistics support for Army missile systems. Identify missile platform life cycle metrics, such as Materiel Availability (Am), Operational Availability (Ao), sparing, cost, maintenance man-hours, and other KSA's, to be optimized by AI. Provide to a logistics engineer knobs to turn to see effects on metrics such as system reliability, system availability, system downtime, administrative delay time, maintenance man-hours, manpower, and OPTEMPO. Additionally, consider a Material Availability (Am) model to also include full life-cycle and fleet-wide sustainment concerns such as fielding schedules, recaps, resets, demils, software and hardware upgrades, modernizations, etc.SoSAT is a government-owned software package and will be provided. Notional and/or representative Army missile reliability and supply data will be used. The size of the dataset will also be representative of actual datasets used and expected to be used by future Hypersonics systems -- a typical 30-year Army sustainment model is approximately 25GB, and multiple models could be combined, yielding datasets in the range of 100-200GB. Any AI solution will need to run on US-government network computers and will be export controlled.

PHASE I: Perform a design study to determine how to use artificial intelligence, machine learning, and real-time computational intelligence to optimize sustainment and logistics support. Deliver a final design of AI's capabilities, a simulation model of Army missile assets (including systems of systems), and a demonstration of an AI-infused logistics model capable of making intelligent trade-off decisions to meet specified PM threshold and objective sustainment metrics -- specifically, downtime and readiness levels as calculated by Army missile systems, using inputs such as failure rates, ALDT, repair times, and maintenance man hours. A successful design will be able to optimize support, minimizing missile system downtime and maximizing system availability, using logistics inputs (component failure rates, repair part shipping times, repair times, maintenance man hours and maintainer staffing). Designed AI must be capable of handling, learning from, living in, and analyzing datasets upwards of 200GB in size. Designed AI must also show a 75% reduction in results data processing time over current methods, a 10% reduction in data input, import, and formatting time over current methods, and a 30% reduction in output dataset size. Test method to determine success for above metrics will be accomplished through analysis.

PHASE II: Deliver and implement a working prototype of an AI-infused logistics model (as designed in Phase I) capable of deep learning and making intelligent trade-off decisions to meet specified PM threshold and objective sustainment metrics. The model will also provide the capability to measure the impacts of technology insertions, obsolescence, reset, and other significant events in the entire Army missile platform's life cycle, and to optimize such downtime and upgrade scheduling over that typical life cycle. Prototype AI must be capable of handling datasets upwards of 200GB in size. Prototype AI must be able to learn from baseline sustainment datasets, learn from excursion datasets on the fly, and apply learned behaviors. Prototype AI must show a 100% reduction in results data processing time over current methods, a 20% decrease in data input, import, and formatting time over current methods, and a 50% reduction in output dataset size. Test method to determine success for above metrics will be accomplished through demonstration. Mission profiles and operations in the model will be based on notional Army missile concept of operations (CONOPS).

PHASE III: Deliver a polished and complete working AI-infused logistics sustainment model making intelligent trade-off decisions to meet specified PM threshold and objective sustainment metrics to all Army PM's and for current and future Army missile platforms. The final product should model and optimize logistics and sustainment at multiple levels of fidelity from battalions to component parts, from components to systems of systems, from individual missions to entire life cycles, use advanced web and cloud services to compute and be hardware-independent, may include an asynchronous mobile application to view and sort results, handle upwards of 1TB of data, and be hosted or otherwise available to all CAC-enabled personnel. Test method to determine success for above metrics will be accomplished through operations.

KEYWORDS: artificial intelligence, logistics, simulation, modeling and simulation, sustainment, availability, reliability, maintainability, supportability, software development, machine learning, neural networks, real-time computational intelligence, data science, software architecture, deep learning, support vector machines, Levenburg-Marquardt, particle swarm optimization

References:

Rosienkiewicz, Maria. (2013). “Artificial Intelligence Methods in Spare Parts Demand Forecasting. Logistics and Transport.” 2013.; Real Carbonneau, Kevin Laframboise, Rustam Vahidov. “Application of Machine Learning Techniques for Supply Chain Demand Forecasting.” European Journal of Operational Research, Vol. 184, Issue 3, 1 Feb 2008, pp. 1140-1154.; C. Jennings, D. Wu and J. Terpenny, "Forecasting Obsolescence Risk and Product Life Cycle With Machine Learning," in IEEE Transactions on Components, Packaging and Manufacturing Technology, vol. 6, no. 9, pp. 1428-1439, Sept. 2016.; Bellochio, Andrew T. (2018). “A Framework to Enable Rotorcraft Maintenance Free Operating Periods.” 2018.; Pukish, M.S.; Reiner, P.; Xing Wu, “Recent advances in the application of real-time computational intelligence to industrial electronics,” IECON 2012 – 38th Annual Conference on IEEE Industrial Electronics Society, pp. 6305, 6314, 25-28 Oct. 2012

US Flag An Official Website of the United States Government