Seminar za primenjenu matematiku, 6. decembar 2011.

• 05. Decembar,  2011
• Komentari (0)

Naredni sastanak seminara za primenjenu matematiku održaće se u utorak, 06.12.2011. u 14:15, u sali 301f, MI SANU.

Predavač: Rastko Selmić, AT&T Associate Professor of Electrical Engineering, Louisiana Tech  University Ruston, Louisiana, USA

Naslov predavanja: FAULT DETECTION IN NETWORKED DYNAMICAL SYSTEMS

Sadržaj: We present a dynamical model of wireless sensor networks and its application to sensor node failure detection. Recurrent neural networks are used to model a sensor node, its dynamics, and interconnections with other nodes. The modeling approach is used for sensor node identification and fault detection. The input to the neural network is chosen to include delayed output samples of the modeling sensor node and the current and previous output samples of neighboring sensors. The input to the neural network and topology of the network are based on a general nonlinear dynamic sensor model. We also discuss a leader-follower networks model and a neural net-based observer for such dynamical network. The observer is used in propagation analysis of a fault through the network. We introduce a fault at the power-leader and study its detection at the leader and followers. It is shown how such leader-follower network model can be used in human emotions modeling for a network of people. To verify and validate the approach, we simulate the individual nodes, the network, and the network observer using Simulink that allows for intuitive and visual networking of dynamical elements.

---