Study On Composition Selection And Efficiency Increase Of Hot Rolled Thread Steel Based On Intelligent Algorithms

In this paper,the appearance that the change of trace element ratio in rebar can significantly change its mechanical properties is studied.Through the processing of a large number of actual production data of a steel enterprise,statistical methods and intelligent algorithms are applied to explore the dependence between the mechanical properties of the rebar and the composition of trace elements.According the data of all trace elements in rebar and its mechanical properties(yield strength,tensile strength and elongation after fracture),the correlation analysis and principal component analysis are carried out,using statistical model,combining theoretical analysis with actual production test,an optimized solution is explored to reduce the Mn element content ratio under the condition of unchanged Cr content and guarantee the mechanical properties of the rebar.The aim is to provide a theoretical reference for the efficiency of steel companies.This study mainly draws the following conclusions:1.In order to realize the change of micro-alloy content of rebar,a stepwise multiple linear regression model is applied to the main components C,Mn,S,P,Si,Cr,Ni and V which affect the mechanical properties,a statistical linear dependence between the main component content of the rebar and its mechanical properties tensile is established.The theory of statistical linear model is used to guide the actual production test,which shows that it can improve the production efficiency of steel enterprise rebar to some extent.Because of the neglect of interaction and nonlinear action between elements,the relationship between the mechanical properties of the rebar and the composition content is unstable,so the stable production target is not achieved.In order to explore the stable relationship between mechanical properties and compositional content of rebar,the inter-element interaction effects and single-element nonlinear effects are studied,and the interaction-free quadratic model,non-interacting cubic model,and interacting cubic model are constructed,and the validation of the results is significantly better than the statistical linear model.In addition,in order to increase the application flexibility of statistical nonlinear models,a multivariate adaptive regression spline model is introduced and good application results are obtained.In practice,the statistical nonlinear model has very significant advantages when it is consistent with the assumptions,however,once the initial conditions have changed or fine adjustments in production parameters,the instability generated in the model,and the expression of the statistical nonlinear model is more complicated,which brings unnecessary trouble for the model correction.2.Based on the research of statistical linear model and nonlinear model,BP neural network algorithm is introduced.This algorithm can improve the training precision by increasing the size of training sample set.It has strong applicability and is used to explore the implicit dependence of the mechanical properties of the rebar and its constituent elements,and the model correction is more convenient.The results of the study show that the error rate of the dependence relationship between the main component content of rebar and its mechanical properties is only 6%,which provides a good guide for the actual production of steel enterprises,and achieves energy reduction and efficiency enhancement to some extent.3.Because the training accuracy of the BP neural network algorithm is influenced by the initial parameter setting,it often takes a long time to obtain the optimal implicit dependency relationship,in order to improve the timeliness of the algorithm,this paper introduces a genetic algorithm(GA-BP)and designs an improved BP neural network algorithm based on genetic algorithm.It retains the BP neural network's ability to use sample learning to achieve arbitrary spatial mapping,and adds the advantages of global search of complex,non-differentiable,and nonlinear functions of genetic algorithms.In the application to explore the relationship between the performance of rebar and its component content,good results have been achieved.The error rate of the dependence relationship between the content of main component elements of rebar and its mechanical properties obtained by this algorithm is only 2.45%,which greatly improves the time efficiency of BP neural network algorithm.4.Based on the theoretical analysis of correlation analysis,principal component analysis,stepwise linear regression,nonlinear regression,MARS modeling,BP neural network modeling and GA-BP neural network,we find that the dependence of the yield strength and tensile strength on the alloying element content is stronger.The dynamic potential of yield strength and tensile strength is studied when the Mn content is within a certain interval under the condition of constant Cr content.The corresponding Mn element can be calculated according to the simulation curve of GA-BP neural network.Through theoretical analysis and actual production test,the reduction in Mn content is finally determined to be 0.1%.