In 2022, the U.S. Department of Defense requested a $3.5 billion budget for depot operations and maintenance which includes critical repairs to aircraft, missiles, aircraft carriers, ships, submarines, combat vehicles, and other equipment. However, the actual approved budget ($2.1 billion) falls far behind what is needed, threatening the readiness of military vehicles. One way to close this gap is to employ predictive and prognostic maintenance (PPMx). The U.S. Department of Defense has already prioritized a shift to PPMx solutions as the technology has been proven to optimize standard maintenance scheduling and reduce time, expense, and downtime due to component failure, repairs, and replacements across industries. This approach continues to grow in its usefulness as more powerful ML techniques and greater amounts of data are collected and able to be easily stored. Now that the technology is mature, there is a need to determine which predictive modeling approaches are best suited to military vehicle maintenance.
This paper details research to identify effective approaches for predictive models for vehicle maintenance utilizing a combination of onboard sensor data, historical and peer vehicle data, and maintainer input for PPMx. This includes a comprehensive review of statistical and ML modeling approaches used for vehicle PPMx, with a focus on anomaly detection, fault classification, and remaining useful life estimates from survival regression models. Finally, a case study is presented demonstrating the way these approaches could be applied to U.S. Army vehicle operations and maintenance. Specifically, opportunities and constraints regarding available data and systems are considered, focusing on technical considerations for a human-in-the-loop approach to integrate maintainer feedback into the predictions and decision support recommendations produced by statistical models. This integrated approach can provide more effective and holistic PPMx solutions to help drive vehicle maintenance decisions, reduce costs, and increase readiness in the U.S. military.
Keywords
ANALYTICS,ARCHITECTURE,DECISION,MACHINE LEARNING,VISUALIZATION
Additional Keywords
PPMx