Data Reduction Techniques for Real-time Fault Detection and Diagnosis, and Multiple Fault Inference with Imperfect Tests

Award Information
Agency: National Aeronautics and Space Administration
Branch: N/A
Contract: NNX08CD30P
Agency Tracking Number: 070120
Amount: $99,968.00
Phase: Phase I
Program: STTR
Awards Year: 2008
Solicitation Year: 2007
Solicitation Topic Code: T1.01
Solicitation Number: N/A
Small Business Information
100 Great Meadow Road, Suite 603, Wethersfield, CT, 06109-2355
DUNS: 808837496
HUBZone Owned: N
Woman Owned: N
Socially and Economically Disadvantaged: N
Principal Investigator
 Sudipto Ghoshal
 Principal Investigator
 (860) 257-8014
Business Contact
 Sudipto Ghoshal
Title: Business Official
Phone: (860) 257-8014
Research Institution
 University of Connecticut
 Not Available
 438 Whitney Road Ext., Unit 1133
Storrs, CT, 06269
 (860) 486-3994
 Domestic nonprofit research organization
The recent advances in data collection and storage capabilities have led to information overload in many applications, including on-line monitoring of spacecraft operations with time series data. Such datasets present new challenges in data analysis, especially for implementation in memory-constrained DECUs. Also, the traditional statistical methods break down partly because of the increase in the number of observations (measurements), but mostly due to an increase in the number of variables associated with each observation ("dimension of the data"). One of the problems with high-dimensional datasets is that not all the measured variables are "important" for understanding the underlying phenomena of interest. In addition to the computational cost, irrelevant features may also cause a reduction in the accuracy of some algorithms. The first key issue we propose to address is that of data reduction techniques for onboard implementation of data-driven classification techniques in memory-constrained onboard processing units. Some of the classification techniques we intend to use with the above data-reduction techniques include, support vector machine (SVM), probabilistic neural network (PNN), k-nearest neighbor (KNN), principal component Bayesian analysis (PCA). To improve the diagnostic accuracy and efficiency of the above classifiers, we will apply classifier fusion techniques such as AdaBoost, Error correcting output codes, Voting to find which architecture will enhance the accuracy and under what conditions. Finally we will investigate Dynamic Multiple Fault Diagnosis that can work with imperfect fault/anomaly detection tests. As part of this task, we will develop novel factorial hidden Markov model-based inferencing techniques such as Lagrangian relaxation and Viterbi decoding algorithms to solve this difficult combinatorial optimization problem, for on-board vehicle health monitoring and fault diagnosis.

* 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