On the Zener–Hollomon Parameter, Multi-Layer Perceptron and Multivariate Polynomials in the Struggle for the Peak and Steady-State Description

• Main Authors: Petr Opěla, Petr Kawulok, Ivo Schindler, Rostislav Kawulok, Stanislav Rusz, Horymír Navrátil
• Format: Article
• Language: English
• Published: MDPI AG 2020-10-01
• Series: Metals
• Subjects: flow stress description; peak and steady-state description; regression analysis; Zener–Hollomon parameter; artificial neural networks; multivariate polynomials
• Online Access: https://www.mdpi.com/2075-4701/10/11/1413

Description of flow stress evolution, specifically an approximation of a set of flow curves acquired under a wide range of thermomechanical conditions, of various materials is often solved via so-called flow stress models. Some of these models are associated with a description of significant flow-curve coordinates. It is clear, the more accurate the coordinates description, the more accurate the assembled model. In the presented research, Zener–Hollomon-based relations, multi-layer perceptron networks and multivariate polynomials are employed to describe the peak and steady-state coordinates of an Invar 36 flow curve dataset. Comparison of the utilized methods in the case of the studied alloy has showed that the suitable description is given by the multivariate polynomials although the Zener–Hollomon and perceptron networks also offer valuable results.