On The Design Of The DISO Controllers For A Neuro-Fuzzy Vector Control. The Training Data Acquisition

The paper deals with the synthesis of some neuro-fuzzy speed and current controllers for a vector control based structure associated with an asynchronous motor and an inverter. Many operating conditions for the system are taken into account, like speed reference profile (including some special ones for a smooth motion in a transportation system), quadrants, load torque etc. The final goal is to substitute all the standard controllers with neuro-fuzzy variants, able to ensure a high robustness so that the influences of the motor parameters, load torque, sampling period and operating conditions could be minimized and good global performance could be achieved. The paper intends to reveal some details not directly visible applying an automatic neural network training procedure for the generating the fuzzy controllers. However, these details could improve the solutions. The results are compared with those from the initial / reference system.