Current USMC logistics information systems do not possess the predictive modeling and simulation tools required to support strategic mission critical MAGTF planning efforts. Standard maintenance and supply information (service requests, spare parts requisitions, NIIN inventories, fleet readiness metrics, etc.) is readily available in an ERP system and is visualized via custom-built asset health/fleet readiness dashboards. The dashboards are dynamic; however, they lack critical scenario planning capability (e.g. deployment to a remote desert environment) by integrating key intelligence data. Tagup is proposing to build and validate risk-based asset survival models on key LAV25 maintenance and supply data. Survival models will be used to estimate probability of failure and model time to event as a function of maintenance status (e.g. deadlined, operational degraded, etc.), cost and failure mode. Potential savings will be identified as a result of model accuracy (as a function of increased asset availability) with a plan to validate model outputs on live streaming data across target USMC functions/users (Phase II).
Benefit: Using machine learning, we have identified a potential savings of ~$77MM (or 5% of the fleet value) for a single TAMCN (LAV25) over the life of the asset by: (1) increasing operational availability by 6.5% and (2) increasing confidence in asset availability by reducing excess inventory management. An increase in LAV25 readiness is a result of improved maintenance productivity (identifying components with the highest probability of failure), optimized equipment procurement/inventory management and a tool that enables improved replace vs. repair decision making. Under conservative assumptions, this scales to $1.15 billion over 236 critical TAMCNs in USMC MAGTF, a small force relative to the USMC Aviation division, broader Department of Navy, USAF, Army and generally DLA/LOGCOM operations.
Keywords: asset management, asset management, Data science, Failure Prediction, Machine Learning, Data Analytics, survival analysis, internet of things