TECHNOLOGY AREA(S): Air Platform, Information Systems

ACQUISITION PROGRAM: PMA-276 H-1 USMC Light/Attack Helicopters

OBJECTIVE: Develop a suite of novel data analysis tools, and the integration of data mining with physics-based models, to quickly assess current rotorcraft diagnostic state, make predictive life analysis, detect and address anomalies, and provide a complete traceability of part history.

DESCRIPTION: Navy aircraft data are stored in several database management systems, both in digital format and paper records. Each of the Navy�s type/model/series aircraft has its own data characteristics that depend on several factors such as (1) aircraft category (i.e., fixed or rotorcraft), (2) installed data recorders and sensors (such as the Integrated Mechanical Diagnostics System or the Vibration, Structural Life, and Engine Diagnostics System), and (3) any unique functional line duties and records that would be needed during maintenance service events (such as remove and replace, service fluids, inspection criteria, etc.). Additionally, the NAVAIR Enterprise Condition Based Maintenance Plus (eCBM+) team encourages solutions that support an open architecture data management and engineering analysis environment.

Specifically, with respect to life management and maintenance, and repair and overhaul (MRO) activities, numerous data are collected and stored at various geographical locations in different management and computer systems. This architecture results in multiplicity in data, some contradictory data, and incomplete data. Conservative engineering judgments are often made to resolve these data inconsistencies when it is difficult or impossible to correct or rebuild all datasets. Many times, aircraft life is penalized for these data discrepancies.

An analysis toolset is needed with a reasoning engine that: interfaces with aircraft and external data systems; can provide current diagnostic state of the aircraft; and is able to make component life predictions. Developing this analysis tool will require describing types of data sets; constructing/adopting necessary standards and metadata; implementing machine learning (ML) algorithms; conducting predictive analysis; and presenting the required data to the end-user in a convenient but familiar and decision-ready format. Pre-processed data to be aggregated will include large maintenance datasets (e.g., part installs/removals, pilot flight reports, inspection records, teardowns, and field tech reports), operator information (e.g., mission types, locations), flight test data (e.g., Health and Usage Management Systems (HUMS)), and engineering data (e.g., design specifications, technical drawings, manuals, failure modes, on-ground tests). New paradigms on data transformation, data mapping, data mining and data visualization should be explored for enhancing the data processing capability of current systems and processes. The data processing should result in useful interfaces including, but not limited to: a current snapshot of the aircraft health since the last inspection; load spectrum development; schedule indicating the next inspection time; updates on component retirement; component replacement prioritization; updates identifying events of interest; event root cause analysis; and risk assessment. The system should be able to identify and extract such useful knowledge from large quantities of data for making informed decisions on aircraft state and its components. Resilience to both data and processing faults is sought as faults can cause data corruption and can have many different sources due to software bugs and hardware errors. The analysis toolset needs to be: (1) able to handle structured and unstructured data; (2) able to identify and resolve data quality issues; (3) resilient to both data and processing faults; (4) quick (e.g., have a low latency retrieval of data ranging between 24-48 hours depending on criticality of alert or action needed); (5) based on modular, user-friendly, highly-customizable applications that will respond to different functional end-user needs; and (6) easily scalable. Lastly, the analysis toolset should be fully compatible with existing Navy and Marine Corps Intranet (NMCI) and logistics enterprise systems, including but not limited to relational database management systems, open source architecture, Java, Python, web compatibility (e.g., ozone widget framework), and support for Public Key Infrastructure (PKI) certificate login. The solution must meet the system DoD accreditation and certification requirements as cited in DoDI 8510.01, Risk Management Framework (RMF) for DoD Information Technology (IT), and DoDI 8500.01, Cybersecurity [Refs 5, 6].

PHASE I: Design, develop, and demonstrate the feasibility of a data analysis toolset able to meet the requirements outlined in the Description. Ensure that compliance with NMCI, information assurance (IA), and cyber security is being considered throughout planning and development. Develop plans for a prototype to be developed in Phase II.

PHASE II: Further develop the proposed technology to use ML algorithms to aggregate a variety of pre-processed data from multiple sources. Use data mining techniques and usage-based and/or physics-based models to provide useful information (i.e., predictive analysis, load spectrum development, inspection schedule, part updates, events of interest, event root cause analysis, risk assessment) in a convenient, intuitive, and decision-ready format for different functional end-users.

Ensure compliance with NMCI, IA, and cyber security is continuing throughout planning and development. Demonstrate the prototype analysis toolset in an isolated yet representative operational environment.

PHASE III DUAL USE APPLICATIONS: Transition and integrate the data analysis toolset into the Navy logistics enterprise system to be used with actual flight and fleet maintenance data. Perform necessary IA and software qualification testing to be able to operate within NMCI environment. Validate the production system functionality for Navy/Marine rotorcraft and/or fixed wing program of record. Provide deployment and training to user base community, including user manuals and functional guides.

Successful technology development would benefit the data analysis industry as a whole, providing the private sector with tools to perform quality assurance, sort, reduce, transform, display, and make projections on multiple large datasets. Potential areas that can benefit include engine manufacturers; energy production, automobile, and medical industries; and the Department of Health and Human Services.

REFERENCES:

1. Bharadwaj, R., Mylaraswamy, D., Vechart, A., Smith, M., Figliozzi, P., Biswas, G., & Mack, D. �Case Studies: Use of Big Data for Condition Monitoring�. AIAC 16 Sixteenth Australian International Aerospace Conference, Melbourne, Australia, 23-26 February 2015. www.humsconference.com.au/Papers2015/Non_peer_Reviewed/Vechart01.pdf

2. Koelemay, M. & Sulcs, P. �Leveraging Massively Scalable Data Analytics Technologies to Enable Rapid HUMS-Based Fleet Management Decision Support�. AHS 72nd Annual Forum, West Palm Beach, Florida, USA, May 17-19, 2016. https://vtol.org/store/product/leveraging-massively-scalable-data-analytics-technologies-to-enable-rapid-humsbased-fleet-management-decision-support-11477.cfm

3. Shaw, J. �Why �Big Data� is a Big Deal�. Harvard Magazine, March-April 2014. https://harvardmagazine.com/2014/03/why-big-data-is-a-big-deal

4. Gavrilovski, A., Jimenez, H., Mavris, D., Rao, A., Marais, K., Shin, S., & Hwang, I.� �Challenges and Opportunities in Flight Data Mining: A Review of the State of the Art�. AIAA Infotech @ Aerospace San Diego, California, 2016. https://arc.aiaa.org/doi/pdf/10.2514/6.2016-0923

5. DoDI 8510.01, Risk Management Framework (RMF) for DoD Information Technology (IT), dated 12 March 2014.� http://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodi/851001_2014.pdf

6. DoDI 8500.01, Cybersecurity, dated 14 March 2014.� http://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodi/850001_2014.pdf

7. DoDI 8582.01, Security of Unclassified DoD Information on Non-DoD Information Systems, dated June 6, 2012, Change 1, October 27, 2017. http://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodi/858201p.pdf

KEYWORDS: Fatigue Life; Diagnostics; Prognostics; Modeling; Big Data; Machine Learning