Correlation Research And Visualization Analysis Of Bake Data

In the production process of aluminum anodes,it is divided into four processes: calcination,assembly,roasting,and forming.The roasting stage has the greatest impact on the quality of the anode,so it is of great significance to study the roasting process.During the roasting production process,a large amount of time-related roasting process and anode quality time series data will be accumulated.The rational use of these data and the discovery of the correlation between the process and quality parameters have important application value for improving the quality and yield of anode production.In this paper,based on the improved long-term and short-term memory neural network,the correlation and data analysis of roasting data are studied and visualized.It is mainly divided into six steps: raw data preprocessing,feature selection and derivation,model building,model training,visualization of prediction results,correlation analysis,and system implementation.In order to ensure the feasibility of training,it is necessary to reorganize the data in the original format first,and slice and segment according to the time step.In order to eliminate the influence of the dimensions of different parameters on the results,the 0-1 mean method is used to centralize and standardize the original data.Feature extraction is the top priority in data mining.In order to improve the robustness and accuracy of the model,in this paper,according to the actual situation,in the original data features,the difference between the temperature of the firing block before and after the fire passage is added as the slope information of the firing curve,And increase the temperature difference between the same firing block at the same time around the fire passage as a new feature.In this paper,the traditional LSTM is improved according to the characteristics of the roasting time series data.The output ht-1 at the previous time is not allowed to participate in the calculation of the forget gate at the current t.A matrix of all 1s with the same dimension as the dimension of ht-1 and only The matrix including the forgotten gate input with xt at the current moment is concatenated to update the cell memory Ct-1.The improved result makes the left half of cell state C never forget,which can reduce the size of forgetting and reduce the amount of cell state update.This article uses the anode production data from 2012 to 2018 provided by an aluminum plant to build a prediction model and visualize the prediction process and results.The test results show that the model classification prediction accuracy is slightly higher than the traditional LSTM.Then this article uses basic correlation analysis,methods based on improved LSTM algorithm and sensitivity analysis,and complex correlation coefficients to conduct in-depth correlation research and visual analysis of quality factors and temperature process factors.It is of great significance to improve the quality of anode production and the economic benefits of aluminum plants,and save energy.Finally,based on the classification model,this paper develops a Web-based roasting data correlation research and visual analysis system.The system mainly includes system management,data collection,data preprocessing,historical data visualization analysis,model training,quality prediction,and roasting data.Correlation analysis and real-time monitoring of the temperature change of the furnace chamber at the current time and abnormal alarm prompting of the roaster portrait,model comparison and evaluation and other functions.After testing,the system has good performance and stable operation.