NASA Logo, National Aeronautics and Space Administration
Intelligent Systems Division Banner

Presentations on Machine Learning Work Given to ACM

Mark Schwabacher gave a two-hour invited talk to the San Francisco Bay Area Chapter of the Association for Computing Machinery (ACM) Data Mining Special Interest Group (SIG) entitled “Using Supervised and Unsupervised Learning to Detect and Isolate Faults in Rocket Engines.” The talk took place on April 8 at SAP Labs in Palo Alto, and was attended by approximately 20 people.

The talk described two classes of algorithms to automatically detect and isolate faults in rocket engines. Supervised learning algorithms require training data consisting of a large number of labeled examples of sensor data from both nominal operation and from failures. They learn a model that can distinguish among nominal data and data from each failure mode, and can thus perform both fault detection and fault isolation. In real rocket engine sensor data, there are not enough failures to allow supervised learning to be used, so this class of algorithms has only been used with simulated data.

Unsupervised anomaly detection algorithms are trained using only nominal data, learn a model of the nominal data, and signal a failure when future data fails to match the model. They are not able to identify the failure mode, but they can be trained using real data that does not include any failures. Examples of unsupervised anomaly detection algorithms include the Inductive Monitoring System (IMS), Orca, GritBot, and one-class support vector machines.

Schwabacher presented results of applying unsupervised anomaly detection to detecting faults in real data from the Space Shuttle Main Engine, and of applying supervised learning to detecting and isolating faults in simulated data from the J-2X rocket engine.

The talk was videotaped and is available on the Web at: http://fora.tv/2009/04/08/Mark_Schwabacher_Fault_Detection_in_Rocket_Engines.

BACKGROUND: The talk included work performed under four projects over the past five years with collaborators at ARC, SSC, MSFC, and Pratt & Whitney Rocketdyne (formerly Boeing Rocketdyne). All four projects involved using machine learning to detect and isolate faults in rocket engine data.

COLLABORATORS:

  • NASA Ames Research Center: Nikunj Oza, Bryan Matthews (SGT), Ashok Srivastava, and Rodney Martin
  • NASA Stennis Space Center: Fernando Figueroa
  • NASA Marshall Space Flight Center: Anthony Kelley and John Butas
  • Pratt & Whitney Rocketdyne: Robert Aguilar, Matt Davidson, Al Daumann, John Stephens, Chuong Luu and Hagop Panossian

NASA PROGRAM FUNDING: Exploration Systems Mission Directorate Technology Maturation Program as part of the ISHM Testbeds and Prototypes Project; ESMD Advanced Space Technology Program as part of the Collaborative Decision Systems Project; ESMD Exploration Technology Development Program as part of the Integrated Systems Health Management Project; The NASA Innovative Partnerships Program, under a task entitled “Test-Stand and J-2X Engine End-to-End Integrated System Health Management (ISHM) Capability”; Boeing, under a Space Act Agreement; Pratt & Whitney Rocketdyne, under a Space Act Agreement

Contact: Mark Schwabacher

05/07/2009

+NASA Home

+Ames Home

Presentations on Machine Learning Work Given to ACM
+ Home + Organization + News + Research Areas + Publications + Software + Internal Systems Help
First Gov logo
NASA Logo - nasa.gov