Fault classification based upon self organizing feature maps and dynamic principal component analysis for inertial sensor drift
Por:
Benítez-Pérez H., García-Nocetti F., Thompson H.
Publicada:
1 ene 2007
Resumen:
Fault detection and identification is an active research field in several application areas. There are still many challenges in on-line detection and identification. Over The years several approaches have been pursued based on model-based or knowledge-based techniques, however, these present several practical drawbacks with regards to time consumption or lack of adaptability. Here a mechanism to classify both previously encoun-tered faults and also new novel faults is presented. This is based upon a combination of a statistical approach, Principal Component Analysis (PCA), and non-supervised neural networks, Self Organizing Maps (SOM). Simulation results are presented through insertion of incipient faults into the inertial sensors of an aircraft flight control system and an evaluation of the proposed approach is made.
|