Active magnetic bearing control system testing and validation using a multiobjective genetic algorithm

An off-line preliminary assessment of an evolutionary controller testing and validation technique intended foil online use is described. This approach to closed-loop performance and stability assessment was proposed following a control system design procedure undertaken for the active magnetic bearing rig housed at Rolls-Royce, Derby. Previous control design work for this rig was conducted online due to discrepancies between the plant and best available model. The controller validation technique being assessed here employed a multiobjective genetic algorithm to generate various disturbance signals. These disturbances were then applied to a simplified plant model of the AMB rig in order to assess the technique´s suitability prior to its use on-line. The closed-loop model is assessed for stability according to it´s time responses alone as would happen in the on-line case