Determination of DBTT of Functionally Graded Steels Using Artificial Intelligence

This study applied three Artificial Intelligence (AI) models to project the Ductile to the Brittle Transition Temperature (DBTT) of Functionally Graded Steels (FGS). These prediction models are Minimax Probability Machine Regression (MPMR) model, Genetic Programming (GP), and Emotional Neural Network (ENN) algorithms with strong prediction performance. The data of FGS type, crack tip configuration, the thickness of the graded ferritic zone, the thickness of the graded austenitic region, the distance of the notch from the Bainite or Martensite intermediate layer, and temperature were used as inputs in the establishment of the AI ​​models. Charpy impact test (CVN) values obtained from experiments used as output. The datasets have been divided into two groups: one for training and another for testing. The performance of the established AI models was evaluated through 16 statistical indicators and graphically used regression error characteristics, an area over the curve, Taylor diagrams, and scatter plots. As a result, the GP model showed superior prediction performance to other models. The primary objective of this study was to decrease the parameter count while also facilitating model comparisons. In this way, in areas with complex studies such as civil engineering; it allows the work to be completed more practically.