Research On Process Window Design Method For Customized Production Of Plates Based On Property Prediction

The mechanical properties of plate products are part of the items that customers focus on.As the market demand for plates gradually shifts to the mode of customized small batches,the mechanical property requirements of different orders are obviously different,and enterprises are faced with the realistic need to enhance the ability to design flexible processing parameters.Property prediction and parameter design can reduce the production cost of plate products,shorten the delivery cycle,and improve the comprehensive competitiveness,which is of great significance for improving the level of product performance and quality control.However,the current data-driven property prediction methods do not consider the evolution of microstructure and the influence of parametric process capability during rolling.As a result,when using the prediction results to design the process parameters,the formed process window is not reasonable,and it is difficult to implement in the production process.In this dissertation,the methods of microstructure behavior analysis,feature selection of process parameters,multi-output property prediction modeling,parameter processing capability analysis,and process parameter design optimization are deeply studied.Based on the theoretical relationship of "process parameters-microstructure-mechanical properties" of plate products and massive actual production data,a processing feature extraction method based on rolling behavior calculation,a multi-output mechanical property prediction modeling method based on rolling behavior feature fusion,and a customized production process window design method with multi-property target for plates are proposed.By integrating the above methods,a customized production process window design and optimization system with multi-property target for plate products is developed.The main research contents and results are as follows:(1)Aiming at the problems of multi-source heterogeneity,high dimension,nonlinear,strong correlation and observation data cannot reflect the evolution behavior of the rolling process in the property prediction modeling data of plate products,a processing feature extraction method based on rolling behavior calculation was proposed.The method is based on a physical metallurgical model,which maps part of the processing information from the original process parameter space to the microstructure space.The actual state of each rolling pass is effectively characterized by using the information of austenite grain size change in the rolling process,and a feasible theoretical verification method is provided for the subsequent parameter design.Then,based on the maximum mutual information coefficient method,the correlation between each rolling behavior feature,actual process parameters and property indexes is calculated,which eliminates the influence of a large number of irrelevant parameters and provides valuable data for mechanical property prediction modeling.(2)Aiming at the problems of insufficient accuracy of traditional theoretical property prediction methods,too many input parameters for machine learning methods,and difficulty in ensuring the fairness and information integrity of the mapping relationship between parameters and property indicators for single-output property prediction modeling methods,a multi-output support vector regression property prediction modeling method based on the feature fusion of rolling behavior feature and process parameter is proposed.The method utilizes the ability of rolling behavior features to characterize the machining process,and replaces 81 original production process parameters with 20 rolling behavior features in the modeling input parameters.And reduce the overall average relative error of performance prediction from 3.03%to 2.83%.It achieves higher prediction accuracy with fewer input parameters,has a simpler model structure than other property prediction methods,and is more robust under small sample conditions.(3)Aiming at the processing parameter design requirements of customized production of plate products,a manufacturing process window design method for customized production with multi-property targets and the parametric processing capability taking into account is proposed.The method is based on the improved sensitivity weighted control particle beam optimization algorithm,and by introducing parameter processing capability constraints,the design problem of parameter values in the processing process is transformed into the interval design problem of the process window.And the rapid design of process parameters in the production process with the goal of product qualification rate is realized.The experimental results show that the design model proposed in this dissertation can quickly and accurately solve the process parameter processing window that simultaneously meets multiple mechanical property targets.Compared with other traditional parameter design models,it has a more feasible process parameter window and a higher level of pass rate.(4)Aiming at the needs of an steel enterprise to improve the customization and flexible production capacity of plate products,a process parameter design and optimization system for customized production of multi-property objectives has been designed and developed,and it has been practically applied in the enterprise.The system integrates the methods studied in this dissertation,such as feature selection and fusion,parameter processing capability analysis,property prediction modeling,sensitivity analysis,and process parameter design optimization algorithms.Based on the actual production data foundation of the whole-process quality big data analysis platform,the system solves the problem of designing and optimizing the processing window of the production process parameters for the customized requirements of multiple property indicators.The System can help enterprises to develop and analyze new products,help realize flexible and customized production,improve the control level mechanical properties of plates,and enhance the comprehensive competitiveness of enterprises.