Master’s thesis defense of Iury Melo Américo: Prediction of Steel Flow Curves Using Artificial Intelligence Algorithms.
On 17 December 2025, at 9:00 a.m., we held the master’s thesis defense of the student Iury Melo Américo, whose thesis was entitled: Prediction of Steel Flow Curves Using Artificial Intelligence Algorithms..
This work evaluates the performance of different machine learning algorithms for predicting metal flow curves using chemical composition, temperature, strain, and strain rate as input variables. The goal is to provide an alternative to laboratory experimentation, reducing the costs of generating these curves, which are essential in the analysis of the industrial metal forming process. The research is classified as applied technological, employing the implementation of a computational prototype divided into two stages: selection and evaluation of regression models, and the development of a web tool for use by engineers. Cross-validation methods, cleaning, and standardization were applied to the dataset extracted from QForm simulations, as well as hyperparameter optimization using libraries such as scikit-learn, TensorFlow, and Optuna. The main models evaluated were Random Forest, XGBoost, Gradient Boosting, SVR, and shallow and deep neural networks. The XGBoost model presented the best performance, surpassing even deep neural networks, with a mean absolute error of 7.53 MPa on the validation set, demonstrating robustness even for materials not present in the training data. It was found that both temperature and strain rate can influence performance, especially in high-alloy steels, suggesting potential topics for future studies. Finally, an API and a web application were developed to facilitate the practical application of the models in an industrial environment. The results confirm that machine learning techniques are capable of supporting the reliable prediction of flow curves, meeting the proposed objectives.
