Hybrid modelling of dynamic softening using modified Avrami kinetics under Gaussian processes

This paper presents a new method of modelling that combines several approaches to anticipate the softening of nickel-niobium alloys during dynamic recrystallization (DRX). The study employs an extensive dataset obtained from hot torsion deformation tests conducted on high-purity nickel and six nickel-niobium alloys. The niobium concentration in these alloys varies from 0.01 to 10 wt % (Matougui et al., 2013). The hybrid technique integrates the Avrami model to provide early predictions about the kinetics of recrystallization and then uses mechanistic modelling to assess the progression of softening caused by dynamic recrystallization (DRX). The integrated technique is improved by using Gaussian process regression analysis, which investigates the softening properties and offers useful insights into the effects of niobium additions on dynamic softening behaviour. This unique hybrid framework combines multiple modelling tools to reveal intricate connections impacted by solute addition, therefore enhancing our comprehension of the physical events that take place during the hot deformation of superalloys. The use of empirical, mechanistic, and machine learning methods in this hybrid model provides a more thorough and detailed investigation of DRX processes in these alloys. [Display omitted] •Hybrid model combines machine learning and Avrami formalism for recrystallization kinetics.•Zener-Hollomon values increase with Nb content, affecting strain rate sensitivity in Ni-Nb alloys.•Novel approach integrates empirical, mechanistic, and machine learning methods for DRX modelling.•Model provides insights for optimizing industrial superalloy production processes.