Research On Width Defect Tracing And Width Prediction Model For Hot-rolled Strip Steel

The width precision of steel strip is one of the most important indicators to measure its quality.This study takes the 1780 mm hot rolling production line of a domestic steel enterprise as the background,and conducts research on width-related issues of hot-rolled steel strip using traditional data analysis and machine learning methods.The main work of this paper is as follows:(1)In response to the current situation where the width defects of steel strips are mainly traced back through manual analysis by process personnel,a data-driven width defect tracing model is proposed.The model first uses the Kalman filter method to denoise the width data of the entire steel strip,and then uses methods such as Gaussian fitting and peak searching to extract width features and establish width defect classification rules.Finally,combined with the actual production process,a hot-rolled steel strip width defect tracing model is established.The model uses 582width-defective steel strip data for performance testing,and the results show that the classification accuracy of width defects reaches 96.72% and the tracing accuracy of width defects reaches 94.16%,effectively achieving automatic classification and tracing of width defects in hot-rolled steel strips.(2)A DTI-SVR-based optimization model for hot-rolled strip width spread prediction during the finish rolling was proposed.First,an improved DTI model was used to rank the importance of feature variables,and high-contributing features were selected as input variables for SVR.The GSK optimization algorithm was then used to optimize the key parameters of the SVR model.Experimental results showed that the proportion of prediction errors within ±3mm of the model after feature selection reached 96.59%,an improvement of 2.85% compared to that without feature selection.(3)Aiming at the online prediction problem of width in hot-rolled finished strip steel,an OSELM model is used as the benchmark model for online prediction,and the OSELM model is combined with the DTI-SVR-based width spread prediction optimization model proposed in the fourth chapter of this paper to obtain the proposed OSELM-based online prediction model for hot-rolled finished strip steel width.Then,the MAE,RMSE,online testing time of the model,and the testing accuracy within the set error range are used as the evaluation criteria for the model,and compared with the benchmark model.The experimental results show that,compared with the benchmark model,both optimization schemes have increased the proportion of prediction errors within ±3mm by more than 20%,and the proportion of prediction errors within ±5mm by more than 12%,and the prediction time can meet the requirements for online prediction.