Artificial neural network models to predict the response of 55NiTi shape memory alloy under stress and thermal cycles

The present work focused on the development and application of artificial neural network (ANN) models to predict the cyclic behavior of a commercially available, binary Ni49.9 Ti50.1 (at. %) shape memory alloy, also known as 55NiTi (55 wt% Ni). This 55NiTi material is well known in the aerospace industry for design of actuators (which operate under several thermomechanical cycles). Using only the predominant factors that influence the cyclic behavior of the alloy as input variables, simple yet practical models are generated that are able to predict the material responses under several stress magnitudes and hundreds of cycles. A comparison between the predicted temperature–strain responses and available experimental data demonstrated a high level of accuracy of the developed ANN models. It was found that the cyclic response of the 55NiTi alloy was more sensitive to the stress magnitude (in comparison with the effect of the other input parameters considered in the study).