Prediction Of Mechanical Properties And Optimal Design Of Steel Grade For Hot Rolled Strips

The prediction of mechanical properties of hot-rolled strip steel started in the 1970 s.It is a difficult problem in the industry of iron and steel metallurgy.It is a very complicated metallurgical frontier technology,which has broad application prospects and profound scientific significance.Constructing a prediction model of steel mechanical properties with high accuracy and reliability is helpful to optimize the design of the composition and process parameters of existing steel grades and to design new steel grades,so as to improve the performance of strip steel products and reduce production costs.The mechanical properties of hot rolled strip steel are affected by chemical composition,rolling process and other factors,and the interaction mechanism is complex.Therefore,a modeling method that combines production data and metallurgical mechanisms is proposed.Via production data and metallurgical mechanism,the influence each factor is deeply analyzed,and the complex high-dimensional nonlinear problem is divided into several sub-problems.The random forest algorithm is used to select the influencing factors of the mechanical properties of hot-rolled strip steel,and the effective dimension reduction of the prediction model is realized.The isolation forest algorithm is used to clean the abnormal data of the hot-rolled production process,which improved the quality of the modeled data.The generalized additive prediction model for the mechanical properties of hot rolled strip is established,the sub-models corresponding to each component and process factor are fitted by cubic spline function,and the sub-models are calculated and modified by local scoring algorithm,then the prediction models of yield strength,tensile strength and elongation of hot rolled strip are established respectively.Experimental results show that the model has high prediction accuracy and reliability.A novel multi-objective whale optimization algorithm with multi-elites guided is proposed in this paper to solve the above model.The population initialization strategy using opposition-based learning makes the initial solutions closer to the true pareto front.Multiple elite solutions are selected on the pareto front by meshing and the principle of maximum crowding distance,and the elite solutions are cross-mutated.The whale optimization algorithm is improved.The nonlinear convergence factor is used to enhance the exploration of the algorithm in the early stage,and the exploration strategy that combines Gaussian perturbation with opposition-based learning and Archimedes spiral migration strategy are proposed to avoid the problem that the solutions are easily out of bounds in original algorithm.The above improved algorithm is used to guide the non-elite solutions to approach the elite solutions quickly.Compared with the classic algorithms,the proposed algorithm has achieved good results in multiple benchmark functions.According to the prediction model of the mechanical properties,the composition and process parameters of hot rolled micro-alloy steel produced by a domestic hot rolling mill are optimized and the mechanical properties-cost multi-objective optimization design models are established.The multi-objective whale optimization algorithm with multi-elites guided is used to solve the above model,which also has a good performance,effectively improves the mechanical properties of micro-alloyed steel products,reduces production costs,and is conducive to develop higher quality strip steel products.