ANALYSIS OF PREDICTIVE CAPABILITIES OF MULTISCALE PHASE TRANSFORMATION MODELS BASED ON THE NUMERICAL SOLUTION OF HEAT TRANSFER AND DIFFUSION EQUATIONS

摘要

The high quality of steel products used in the transport industry is strongly dependent on the parameters of microstructure created during the thermomechanical treatment. Numerical models of phase transformations based on the solution of diffusion equations presented in this work allow one to determine the correlation between parameters of the technological process, changes of microstructure, and product properties. Consequently, these models can be a useful support of the technology design for manufacturing processes. On the other hand, diffusion-based multiscale models are computationally very expensive. Therefore, two simple single-pointmodels, the Johnson-Mehl-Avrami-Kolmogorov (JMAK) equation and an upgrade of the Leblond model, were used as an alternative for fast calculations. Two industrial processes were selected for testing and validation of the developed multiscale models. The first was manufacturing of dual-phase steel strips and the second was manufacturing of pearlitic steel rails. A finite-element (FE) model was used to simulate temperature changes in the macro scale. Single-point models were solved in each Gauss point of the FE mesh. These models were used to analyze a large number of technological variants and to select those giving the required phase composition in products. The developed diffusion-based models were solved in selected points of the product only. In the micro scale these models simulated the austenite decomposition into ferrite, bainite, martensite, and pearlite. The FE method was used to solve the diffusion equation in austenite grains. The initial and boundary conditions for the diffusion model were determined for local thermodynamic equilibrium using ThermoCalc software. Diffusion-based models were used to simulate the best technological variants selected by the single-point models and to predict advanced parameters of the microstructure (morphology, carbon distribution, distribution of properties). All models were compared with respect to their predictive capabilities and computation times.