A large number of data are produced in industrial production all the time.How to make full use of these data to improve production efficiency is one of the research hotspots at present.However,at present,the application of cold rolling data mainly focuses on the storage and reading of data,and the application of data is insufficient.How to make full use of these production data to optimize the existing control means of sheet shape to improve the quality of sheet shape remains to be studied.Therefore,this thesis takes the shape presetting in the shape control system as the research object.Based on the mass production data in cold rolling production,the intelligent algorithm is used to establish a suitable shape presetting model,so as to achieve a more accurate shape presetting value prediction.The main work of this thesis includes:Firstly,aiming at the problem of insufficient data preprocessing in cold rolling research,the whole process from data extraction,parameter selection,outlier detection,missing value filling and data normalization processing is introduced in detail.According to the existing outlier detection methods,the Grubbs algorithm and the isolated forest algorithm are used to detect and eliminate the outlier of the measured rolling data respectively,and the comparison analysis is carried out.The classification of missing values and various processing methods are studied.After comparison,Lagrange interpolation is used to fill in the missing data.Aiming at the problem of different dimensions of rolling parameters,an appropriate method is selected to normalize the data and map the data to the interval [0,1].Secondly,the priority of presetting control means is analyzed,and the strategy of taking the transverse movement of intermediate roll as the rough adjustment means of plate shape control is developed.After the set value of the transverse movement of intermediate roll is calculated,the optimal presetting value of the bending force of intermediate roll and the bending force of working roll is calculated.Finally,the principle of BP neural network is introduced,and its algorithm structure is analyzed.On this basis,the specific BP neural network prediction process is given and the BP neural network model is established.Through the analysis of the results,it is found that there are defects in the process of predicting the roll bending force.In order to improve the prediction accuracy of BP model and solve the problem of poor stability of BP neural network algorithm,Sparrow optimization algorithm was studied,and SSA-BP roll bending force pre-set value prediction model was established.The collected measured rolling data were classified and related parameter data were extracted.The training parameters of SSABP model were determined by analyzing the structure of input data.By comparing the training results,it can be seen that the optimized model is superior to the single BP neural network model in both convergence speed and algorithm stability. |