Transforming legacy data into a mission-ready future

Challenge

For decades, a U.S. military branch had warehoused transactional data that captured logistics aspects of major weapons systems. Yet the opportunity to use that data for predictive analytics remained elusive due to a combination of data access challenges, antiquated analytics approaches, and a lack of functional knowledge to apply the source data in meaningful ways. 

As the chosen contractor by the military command, we were tasked with demonstrating tangible benefits of real-world predictive analytics capabilities for a major weapons system. Our primary project responsibility was to design, develop, and integrate a predictive maintenance analytics suite prototype that could be applied across tactical to strategic echelons. The project would rely on a military-provided data science environment and the authoritative data sources that existed inside the designated command’s data analytics platform.  

Approach

Over a two-year period, we developed the leading prototype solution for deployment to designated organizations across the military branch in collaboration with the military command. Initially focused on aircraft systems, we used broad domain knowledge of the military branch’s enterprise resource systems, commercial best practices, and operational data application across the business and warfighting domains. We structured an environment that was both capable and cost-efficient, and that could be effectively scaled across the military branch’s large user base inside of their network. The data analytics platform we developed incorporates real-time automated connections to authoritative enterprise data for analysis and presentation, streamlining processes and ensuring data quality. We presented the results of the advanced analytics suite we created using Power BI visualization tools, which exploited platform capabilities and maximized user accessibility to increase the number of authoritative sources. 

Applying machine learning techniques, our team of data science experts generated predictive survival functions for each component and then combined that information with current installation times to predict component additional life and failure probability. We generated additional machine learning algorithms to combine predictions with known maintenance intervals, to forecast ideal candidate aircraft for extended contingency operations. 

Impact

Our work resulted in: 

• A data pipeline with 2,224 authoritative data sources and tables in the command’s data platform
• An auditable model with a high degree of accuracy and performance:
  • Achieving a concordance index of between 85 and 90% for our machine learning model
  • A 93% accuracy of our machine learning model predictions when the client evaluated our results assessing if an aircraft would be “down” the next day 
• Decision analytic tools to forecast maintenance, materiel, supply chain, and transportation requirements based on current reported equipment condition (such as flight hours, rounds fired, and calendar life)
• Tools to project unit-level readiness performance over a specified time period using advanced analytic techniques, including machine learning, across strategic and tactical levels 
• The prototype deployed in aviation with the ability to extend it to ground units and other areas that support strategic materiel and maintenance operations 
• Analytics used to inform materiel response to dynamic mission requirements as part of humanitarian operations

The military branch now has a logical, repeatable, traceable, and automated pipeline in place, with near real-time predictive insights scalable for additional weapon systems and organizations within it. The advanced analytics tools we’ve developed convey a common understanding across tactical to strategic echelons—giving leaders a cohesive visual of projections for readiness, maintenance, supply, and logistics requirements across individual system and components. This provides direct, critical decision analytics tools for leaders in contingency situations by allowing them to make risk-informed, proactive maintenance, supply and operational decisions.