Development and application of a data-driven model for predicting mechanical properties of Ti microalloyed steel
In recent years,microalloyed steel has garnered widespread attention due to its low cost and excellent performance,finding extensive application in numerous industries such as automotive and construction.The prediction of hot-rolled steel properties is a major direction in steel intelligent manufacturing and holds great significance for enhancing the stability of product performance.Consequently,predicting the mechanical properties of microalloyed steel is highly significant for optimizing the production process,improving product quality,and increasing production efficiency.Data preprocessing methods such as clustering,outlier rejection,and linear normalization were developed based on microalloy steel industrial big data,rolling process theory,and statistical principles.Combined with precipitation thermodynamics,the effective Ti content was calculated,and a forward selection-based method was employed for feature screening.The mechanical property prediction of microalloyed steel is established based on random forest,and the particle swarm optimization algorithm is utilized to optimize the hyperparameters of the random forest,thereby achieving high-precision prediction of the mechanical properties of Ti microalloyed steel.The results indicate that the model has a hit rate of 96.8%for yield strength and 98.9%for tensile strength within a relative error of±6%,and the elongation has a hit rate of 97.9%within an absolute error of±4%.It is verified that the developed model conforms to the laws of physical metallurgy and achieves good model accuracy in online prediction in the field.
data preprocessingrandom foresteffective Ti contentparticle swarm optimization algorithmme-chanical properties prediction
万方数据
维普
