Experimental Investigation and Modeling of Creep Curve of Zr–2.5Nb Alloy by Machine Learning Techniques

In the present work, creep tests on Zr–2.5Nb alloy at different stresses and temperatures in the two-phase region were carried out. The three creep regions were very distinct at low stresses and temperatures, whereas the secondary creep regions narrowed down considerably at higher temperatures and stresses. Data obtained from these creep tests were used to simulate the creep curves by multiple linear regression (MLR) and artificial neural network (ANN) modeling. The MLR model was able to predict the primary creep region accurately; however, it over-predicted the secondary creep region. ANN model could simulate all the three creep regions with very high accuracy, where 98% of the creep strain could be predicted within a deviation of ± 5%. Two different creep experiments were conducted to verify the predictability of the proposed models. The results indicate that the ANN technique can be used to predict the creep curves of two-phase alloys.