Research On Prediction Methods Of Continuous Casting Slab Quality Based On Neural Computing

The metallurgical industry is strongly supported by our country,and continuous casting technology is one of the key technologies in the iron and steel metallurgical industry.In the process of continuous casting slab production,due to the influence of molten steel composition,process elements and human operation,the quality defects of continuous casting slab are often caused,such as center segregation,center looseness,center shrinkage hole,center crack,middle crack and so on.This will seriously affect the quality of continuous casting slab products,not conducive to the production of highquality steel.Therefore,accurate prediction of billet quality is of great significance for continuous casting slab production.Based on the continuous casting slab data,the quality prediction of continuous casting slab is carried out,and the main research contents are as follows:Firstly,a data preprocessing model is established for the characteristics of continuous casting slab data with inconsistent distribution range and high dimension.The model includes two parts: data normalization and data dimension reduction.The consistency of continuous casting slab data is improved by linear normalization method.In the dimension reduction part of the data,the high correlation filtering algorithm is used to process the data,which can effectively reduce the dimension of the continuous casting slab data.Secondly,to improve the quality prediction accuracy of continuous casting slab,a billet quality prediction model based on bidirectional long and short term memory neural network is established.This is a prediction model with multiple inputs and single output.The inputs are the influencing factors of continuous casting slab quality and the output is the quality grade of continuous casting slab.The model takes into account the influence of different factors on the quality of continuous casting slab and predicts it effectively.The simulation results show that compared with the three traditional forecasting models,the proposed model has a good advantage in forecasting.Finally,a prediction model based on sparrow optimization algorithm is established to solve the problem of bidirectional long and short term memory neural network hyperparameter selection falling into local optimal.The model uses the sparrow optimization algorithm to optimize four super parameters of bidirectional long and short term memory neural network,and carries out continuous casting slab quality prediction experiment.The simulation results show that compared with the prediction model based on particle swarm optimization algorithm,the proposed model has outstanding advantages in nonlinear approximation performance and convergence speed.Figure 21;Table 11;Reference 54...