To detect and to isolate faults by classification approach the residual-based pattern recognition approach was used. The classification was performed by a radial basis function neural net. The nodes of the net were trained by the unsupervised learning method. If there is no faults than the distribution of feature vectors within regions of clusters is homogenous. As a change in the homogeneity in the feature space have been detected a new fault cluster is introduced. The fault diagnosis was applied to a autonomous mobile vehicle driven by separate DC (Direct Current) motors for each wheel. The possible faults were defects in the sensors, in the actuators, in the vehicle and in the environment. As an application sensor defects in measuring the velocity of the driven wheels and in the motor voltage measurement were simulated.
Diagnosing priori unknown faults by radial basis function neural network
Diagnose vorher unbekannter Fehler mit dem neuronalen Radialbasisfunktion-Netz
2000
5 Seiten, 10 Bilder, 6 Quellen
Conference paper
English
Construction of fuzzy radial basis function neural network model for diagnosing prostate cancer
BASE | 2021
|Automotive engineering | 1991
|