Title :
Performance metrics for fault prognosis of complex systems
Author :
Vachtsevanos, George
Author_Institution :
Sch. of Electr. & Comput. Eng., Georgia Inst. of Technol., Atlanta, GA, USA
Abstract :
This paper presents a methodology for estimating confidence bounds associated with the task of prediction. A data-driven confidence prediction neural network architecture is introduced that accommodates a learning scheme intended to ´shrink´ the uncertainty bounds, as more information becomes available. Performance metrics are discussed and statistical techniques are employed to define confidence bounds; the methodology is applied to a typical shipboard process.
Keywords :
failure analysis; neural nets; radial basis function networks; statistical analysis; uncertainty handling; GRNN; PNN; RBFN; confidence bound statistical techniques; data-driven confidence prediction neural network architecture; failure analysis; fault prognosis performance metrics; general regression neural network; learning scheme; prediction confidence bounds; probabilistic neural network; radial basis function networks; shipboard process; Artificial neural networks; Asset management; Computer architecture; Delay estimation; Drives; Fault detection; Fault diagnosis; Measurement; Neural networks; Uncertainty;
Conference_Titel :
AUTOTESTCON 2003. IEEE Systems Readiness Technology Conference. Proceedings
Print_ISBN :
0-7803-7837-7
DOI :
10.1109/AUTEST.2003.1243597
