DNN and ANFIS Based Modeling and Control of a Thin SMA Wire

In this work, a Dynamic Neural Network (DNN) and an Adaptive Neuro Fuzzy Inference System (ANFIS) based hysteresis modeling and control strategy is proposed for a Shape Memory Alloy (SMA) wire with 0.001" diameter. Such thin SMA wires can achieve a few Hz operating bandwidth and can be used for surface controls in micro UAV (Unmanned Arial Vehicle) applications. Modeling and Control of the SMA wire is a challenging problem because of its dynamic hysteretic behavior. Different modeling and control techniques used in the literature to compensate the effect of the hysteresis. Several attempts have been made recently to use Neural-Network (NN) based modeling of hysteresis via different network structures. In this paper a DNN and an ANFIS based modeling method of the hysteretic effect is proposed and a comparison is made over the modeling error.