Доступ предоставлен для: Guest

ISSN Online: 2377-424X

International Heat Transfer Conference 12
August, 18-23, 2002, Grenoble, France

Stabilization of a neural network-based temperature controller for heat exchangers

Get access (open in a dialog) DOI: 10.1615/IHTC12.4390
6 pages

Аннотация

The stability of temperature controllers for heat exchangers based on dynamic arti cial neural networks is analyzed here. The control system can be represented as a nonlinear map advancing in time, and the spectral radius of the Jacobian of these maps determines the stability of the system. Since ordinary training methods can sometimes drive the network to weights and biases where the corresponding control system is unstable, a modi cation is developed which minimizes the target error and also reduces the spectral radius to assure the stability of the system after completion of the training process. The techniques developed are tested on an experimental heat-exchanger facility where the results are validated.