Neuronal control of an articial swells generation system

Abstract EN

This paper tries a neuronal approach to control a non linear complex system used for the generation of artificial swells.

Artificial Neuronal Networks (ANN), as universal approximator especially for non linear functions, are used in modeling and controlling the swells generator using a non linear model, that of Pierson MOSKOWICH, which represents well the Mediterranean swell behavior.

The proposed structure is composed of two units: the first is rejerred to as Neuronal Supervisor (NS) for priming the swells generation system and the second as Neuronal Control by Internal Model (NCIM) dedicated to command the same system.

So, the latter unit contains an Inverse Model (IM) as a regulator, a Direct Model (DM) as a predictor, and a low-pass filter expected to reduce the modeling errors.

After adopting a judicious choice of the structure parameters, the obtained results have well confirmed the ANN importance as an universal approximator characterized by the learning capability.