Recurrent neural controller applied for two-mass system

In this paper the structure, data processing and adaptation algorithm of neural adaptive controller are described. The topology of the analyzed neural network is similar to the Jordan structure, which contains internal feedback. Fixed parameters of the controller were optimized according to the BAT algorithm. The neural model is applied in a speed control loop of the electrical drive with elastic joint. In order to obtain better damping of state variables oscillations and higher quality of control, applied controller uses two state variables of two-mass system: motor speed and shaft torque. Tests of the designed control structure has been prepared for nominal and disturbed parameters of the plant. Proposed control structure was analyzed in simulations and experiment (prepared using dSPACE1103card).