Method for predicting dynamic re-crystallization fraction of Nb microalloyed steel

Abstract

The invention discloses a method for predicting the dynamic re-crystallization fraction of Nb microalloyed steel, and belongs to the technical field of crossing of steel research and machine learning.According to the method, a data set of Nb microalloyed steel dynamic re-crystallization behaviors is constructed by using experimental data of existing C-Mn-Nb microalloyed steel dynamic re-crystallization type rheological stress; a BP neural network based on Bayesian regularization is used to establish a model among chemical components, process parameters and rheological stress curve characteristics; and through a dynamic re-crystallization fraction mathematical model, high-precision prediction of the dynamic re-crystallization fraction is realized, so that the workload of a single-pass compression experiment and a quenching experiment is obviously reduced, and the efficiency of predicting the dynamic re-crystallization fraction is improved.

Description

technical field [0001] The invention belongs to the intersecting technical field of iron and steel research and machine learning, and in particular relates to a method for predicting the dynamic recrystallization fraction of Nb microalloyed steel. Background technique [0002] The dynamic recrystallization of austenite during high-temperature deformation of Nb microalloyed steel has a great influence on the subsequent phase transformation behavior and final properties of austenite. The study of dynamic recrystallization behavior can provide a basis for formulating the optimal controlled rolling process. . At present, there are two main methods to study the dynamic recrystallization fraction of austenite. One is to use single-pass hot compression experiment and quench at different deformations, and directly conduct metallographic observation to count the recrystallization fraction; Mathematical models of crystallization fractions. The recrystallization behavior can be descr...

AI Technical Summary

Problems solved by technology

However, in the aspect of dynamic recrystallization fraction prediction, machine learning has not yet been applied, and it is of great significance to carry out the work of machine learning in dynamic recrystallization fraction prediction

[0003] By searching the database of the State Intellectual Property Office and the SOOPAT database, the patent CN106053754B authorized a method for predicting the dynamic recrystallization fraction of high-alloyed materials under time-varying conditions, which obtained the metallographic microstructure through a large number of thermal simulation experiments and quenching experiments, and The traditional dynamic recrystallization kinetic model is improved to a new model that can predict the dynamic recrystallization fraction under time-varying conditions, but this patent has two disadvantages: ① it requires a large number of thermal simulations and quenching experiments, ② the recrystallization fraction The model is only applicable under certain composition and process conditions

Patent CN110068507A discloses a method for correcting the traditional recrystallization model, which obtains samples under different deformation conditions through a large number of physical simulation tests, calculates the recrystallization fraction of the core, and uses numerical simulation software to obtain the true strain of the core of the sample; The fitting method is used to obtain the model parameters, and the obtained model parameters are fitted with the Zener-Hollomon parameters to obtain the recrystallization model parameters under other conditions, but this method has disadvantages: ①The accuracy of the model parameters obtained by fitting is not high, ②It requires a large amount of thermal deformation experiments and numerical simulation experiments, the cost is high and the simulation time is long

Images