Volume 1, Issue 1
Optimization of Stability of Retained Austenite in TRIP-Aided Steel Using Data-Driven Models and Multi-Objective Genetic Algorithms
The important parameters that determine the properties of transformation-induced-plasticity (TRIP) -aided steel are the amount of retained austenite phase present in its initial microstructure and its stability. A large value of carbon equivalent leads to a high amount of retained austenite in the initial microstructure of these steels at room temperature. Looking at it from another angle, a high value of carbon equivalent is undesirable, as it adversely affects the weldability. In this study, we have attempted to resolve this conflict by bringing in the notion of Pareto optimality. Through an evolutionary neural network that evolved through multi-objective genetic algorithms, data-driven models were constructed for both carbon equivalent and fraction transformed. The effect of individual variables on the extent of austenite transformation as inferred by the model was found to be consistent with the principles of physical metallurgy of TRIP-aided steel. Next, using a predator–prey genetic algorithm, a bi-objective optimization task was conducted for simultaneous minimization of carbon equivalent and the extent of transformation. The resulting Pareto frontier was carefully analyzed, and the transformation behavior of different TRIP-assisted steels was also predicted for different straining conditions. Further need for optimizing the heat-treatment schedule is highlighted through selective experimentation.