Integrated Learning Algorithms with Bayesian Optimization for Mild Steel Mechanical Properties Prediction

Abstract

To address the difficulties of evaluating the mechanical properties of energy-intensive mild steel materials during large plastic deformation in seismic engineering, a method for predicting their mechanical characteristics based on intelligent integration technology is provided. The collected experimental data is predicted and analyzed by intelligent technology; the experiment is designed as a two-layer model, with the first layer model employing the random forest (RF) algorithm based on Randomized Bayesian Optimization and the natural gradient boost (NGBoost) algorithm serving as the basic learner. The second layer calculates fusion integration using the findings of the first layer's analysis and single-layer logistic regression. The new fusion integration model reflects the experimental test set more accurately. The link between the stress and strain change trend, change rate, and change value The results indicate that the intelligent integration technology has a fitting impact that is 31.2\% and 29.7\% more than the single RF algorithm and the NGBoost algorithm prediction technologies, respectively. The proposed method is appropriate for assessing massive plastic deformations of mild steel materials under various vertical step angles. The reference value of changes in mechanical properties over time is significant.