PSI - Issue 72

Oleh Yasniy et al. / Procedia Structural Integrity 72 (2025) 188–194

193

Fig. 5. Experimental and predicted hysteresis loop for cycle 1025 at a frequency of 1.0 Hz.

Fig. 6 shows the hysteresis loop for 1025 th SMA loading-unloading cycles constructed using ANN and the corresponding experimental data at a frequency of 5.0 Hz.

Fig. 6. Experimental and predicted hysteresis loop for cycle 1025 at a frequency of 5.0 Hz.

Hysteresis loop plots show how closely the predicted values match the experimental data. This makes it possible to visually assess the accuracy of the model predictions for a cycle that was unseen during the training. The results indicate the ANN model high accuracy in predicting the strain  under the influence of stress  , considering the loading-unloading cycle of the SMA material and the loading frequency f . 4. Conclusions In this work, an artificial neural network of the MLP type was used to predict the hysteresis behavior of NiTi SMAs at different loading frequencies. The results showed that the MLP model can accurately predict the material strain  depending on the applied stress  , cycle number N , and loading cycle frequency f . The low values of the