Distributed Data Mining for Aircraft Health Management

Award Information
Agency: National Aeronautics and Space Administration
Branch: N/A
Contract: NNX12CA02C
Agency Tracking Number: 105253
Amount: $749,997.00
Phase: Phase II
Program: SBIR
Awards Year: 2012
Solicitation Year: 2010
Solicitation Topic Code: A1.15
Solicitation Number: N/A
Small Business Information
CA, Palo Alto, CA, 94306-2937
DUNS: 194937939
HUBZone Owned: N
Woman Owned: N
Socially and Economically Disadvantaged: N
Principal Investigator
 Dimitry Gorinevsky
 Principal Investigator
 (650) 400-3172
Business Contact
 Dimitry Gorinevsky
Title: Business Official
Phone: (650) 400-3172
Email: dimitry@mitekan.com
Research Institution
Aircraft Flight Operations Quality Assurance (FOQA) programs are implemented by most of the aircraft operators. Vast amounts of FOQA data are distributed between many computers, organizations, and geographic locations. This project develops methodology for transforming such distributed data into actionable knowledge in application to aircraft health management from the vehicle level to the fleet level to the national level. The distributed data processing methodology provably obtains the same results as would be obtained if the data could be centralized. The data mining methods are efficient and scalable so that they can return results quickly for 10Tb of distributed data.This data mining technology that we call Distributed Fleet Monitoring (DFM) developed in SBIR Phase I satisfies these requirements. The data are transformed into models, trends, and anomalies. The model training and anomaly monitoring are formulated as convex optimization and decision problems. The optimization agents are distributed over networked computers and are integrated through remote connection interface in a scalable open grid computing framework. Though the data and the computations are distributed, they yield provably the same optimal solution that would be obtained by a centralized optimization.DFM feasibility was demonstrated in the problem of monitoring aircraft flight performance from fleet data using large realistic simulated datasets. We demonstrated efficient computation of quadratic optimal solution by interacting distributed agents. The feasibility demonstration successfully recovered aircraft performance anomalies that are well below the level of the natural variation in the data and are not directly visible. The algorithms are very efficient and scalable. Phase I demonstration extrapolates to processing 10Tb of raw FOQA data in under an hour to detect anomalous units, abnormal flights, and compute predictive trends.

* Information listed above is at the time of submission. *

Agency Micro-sites

SBA logo
Department of Agriculture logo
Department of Commerce logo
Department of Defense logo
Department of Education logo
Department of Energy logo
Department of Health and Human Services logo
Department of Homeland Security logo
Department of Transportation logo
Environmental Protection Agency logo
National Aeronautics and Space Administration logo
National Science Foundation logo
US Flag An Official Website of the United States Government