Methodology for metamodelling of microstructure evolution: precipitation kinetic case study

Abstract

Numerical modeling is an important tool assisting in the designing and optimization of the production technology. The highest predictive capabilities are offered by multiscale modeling. The most important limitation of its wide application is computational cost. One of possible solutions is application of metamodels for fine scale modeling. In this paper, a systematic approach to development of metamodels is presented. All necessary steps, analyzing the model, selecting the metamodel inputs and outputs, gathering the training and testing datasets, choosing a metamodelling technique, training and testing the metamodel are described with a scientific background and practical examples. Development of the exemplary metamodel, replacing thermodynamic modeling of precipitation kinetic is presented.

Machine learning in physico-chemical processes metamodeling

“…xx -вектор ковариаций. Метод широко применяется в задачах молекулярной механики [8], механики жидкости и газа [9], в том числе и реагирующих систем [10], процессов термомеханической обработки сплавов [11]. Метод очень гибок (из-за широкого набора возможных функций корреляции), хорошо работает на небольшой обучающей выборке, но в то же время очень чувствителен к зашумлённым данным, а также требует значительных вычислительных затрат на обучение [1].…”

Metamodelling the Hot Deformation Behaviour of Titanium Alloys Using a Mean-Field Approach

“…Although mean-field models are a suitable tool for describing and predicting the microstructure evolution during thermomechanical treatments, their complex rate equations and connections between the internal variables can limit the application of these models due to the excessive computational time to simulate the microstructural and stress changes during deformation. A reduction in the simulation time while maintaining the accuracy of the predictions can be achieved using metamodelling [8,9]. The workflow for developing a metamodel (also known as a surrogate model) consists of: I) analysing the inputs and outputs from the model of interest, II) defining the metamodelling technique and selecting the necessary metamodel inputs, III) running model simulations to generate a dataset that is later separated into training and test datasets; IV) finally, the metamodel should be trained and tested.…”