Abstract

The comparison of four control schemes based on neural networks for teaching purposes is studied in this paper. The inverse model, inverse model loop, direct scheme with and without models are investigated and compared. The study is based on computer simulations of linear and nonlinear single-input single-output (SISO) typical plants and take into account different aspects such as the training period, the training method (on-line, off-line, simple or with epochs, etc.) network configuration and training signals amongst others. This set of experiences can be used in an advanced control course in the Electrical Engineering curricula, to teach different aspects of neural control.