|
|
|
|
Neural Network-Based Recognition and Diagnosis of Safety-Critical Events |
|
|
|
Performed by :
National Aerospace Laboratory NLR, The Netherlands and
Foundation for Neural Networks SNN, University of Nijmegen, The Netherlands
Project leader: Sybert Stroeve, NLR
|
|
|
|
|
|
|
The objective of this research is to investigate the feasibility of a neural network-based system for automatic recognition and diagnosis of safety-critical non-nominal events in ATM. Such a system is intended to support the safety management of routine operations
- by providing a more complete overview of non-nominal events than is obtained by current means, and,
- by giving a diagnosis of the non-nominal situation via an overview of factors that mostly determine the non-nominal situation.
|
|
|
Neural networks are a powerful and general class of non-linear mappings. A neural network is trained on the basis of a data set with input-output examples. The neural network recognition and diagnosis system will be trained to recognise and diagnose non-nominal events from ATM observable data (see figure below). The training data may, e.g., be based on Monte Carlo simulation data or incident/accident data.
|
|
|
|
|
|
The research is structured in the following work packages |
- WP0 : Project management
- WP1 : Requirements and data sources for neural network-based recognition and diagnosis
- WP2 : Evaluation of data sources for neural network training
- WP3 : Definition of air traffic scenario for neural network application
- WP4 : Methods for neural network-based recognition and diagnosis
- WP5 : Final report
|
|
|
|
| |
|
 |
|
Last validation: 15/03/2007
|
|
|
|
|
|
|
|
