Predicting flow stress of Ni steel based on machine learning algorithm

Abstract

This article builds a stress–strain prediction model based on production data from the steel industry by using machine learning algorithms. Based on the stress–strain data of 9Ni steel hot deformation behavior, the prediction model of flow stress constitutive equation of 9Ni steel is established. Four models, including Arrhenius-type model considering strain compensation, Arrhenius-type model of Stochastic Configuration Networks (SCNs) neural network, Arrhenius-type model of Multi-objective Particle Swarm Optimization (AMPSO) and Support Vector Machine (SVM) model, are adopted in this research. The results show that the Arrhenius-type model considering strain compensation can predict the stress trend under different deformation conditions, but its prediction accuracy has some deviation. The prediction model based on SVM algorithm has the best prediction accuracy. The square of Correlation Coefficient (R2), the Average Absolute Relative Error (AARE), and Mean Square Error (MSE) are 0.99996, 0.002455, and 0.1998, respectively. Based on the data of 9Ni steel hot deformation behavior, the prediction models of machine learning algorithm have good application prospects in steel industry.

Application of Constitutive Models and Machine Learning Models to Predict the Elevated Temperature Flow Behavior of TiAl Alloy

“…Recently, more and more date-driven models such as artificial neural networks (ANNs) [17][18][19], support vector machines (SVMs) [20], random forests (RFs) [21], and Gaussian process regressors (GPRs) [22] have been developed to predict the hot-deformation behaviors of alloys with the development of machine learning techniques. Ge et al [17] utilized the ANN model and Arrhenius type model to predict the hot-deformation behavior of a high-Nb-containing TiAl alloy with β + γ phases.…”