Research On Bar Profile Prediction And Parametric Design System Of The Disc Refiner Plate

As one of the important fundamental raw material industries,the pulping and papermaking industry has the largest proportion in the light industries.The disc refiner is the core equipment in pulping and papermaking,and the refining plate is the "heart" of the disc refiner.At present,refining plate research and development depends on the experiences of a few engineers,which needs continuous adjustment,test,and long cycle.The fact that pulp performance is difficult to control hinders its quality improvement.Previous studies mainly focused on refining intensity,bar force and pulp flowing in the refining zone,the influence of few bar parameters on the pulp performance as well as the influence of the pulp performance on the paper property,which failed to solve the above problems completely.Therefore,this dissertation which is supported by the key research and development program of Shaanxi province(2020GY-105),discussing from the whole process of the raw pulp-bar-pulped-paper,with experiment,detection and analysis,mathematical modeling and practice verification,focuses on the following work:(1)The characteristic selections are carried out for the independent variable of bar profile and the dependent variable of pulp performance,the refining intensity model of SCSL(Specific Contact Surface Load)is established based on the bar contact surface.A hierarchical model is constructed with the AHP(Analytic Hierarchy Process)method,to evaluate the refining plate design.Combined with the refining intensity,the calculation results of the hierarchical model show that the bar width,groove width,bar angle,inner and outer diameter are the main factors in refining plate design.At the same time,the validity of SCSL is verified by experiments.The literature research method is used to summarize the national standard and industry standard of pulp.The field investigation method is used in the actual production of enterprises,and then the characteristic selection of the dependent variable of pulp performance is carried out with Delphi Method,combined with the research results of the evaluation of refining plate scheme.It can be concluded that the main variables of the pulp performance include the beating degree,fiber mean length,tear index,grammage,tensile index and bulk etc.(2)The conception and mathematical modelof the sequential growth rate about the pulp performance are proposed.Based on the experiment research of the influence of the bar on the pulp performance,regression analyses are completed for the main parameters of the bar,bar width ratio to groove and bar contact surface to the sequential growth rate of pulp performance.The analysis results show that the bar main parameters have effects on the pulp performance,but the influence degree is different.The bar width has the greatest influence on the sequential growth rates of beating degree and bulk.The groove width has the greatest influence on the fiber mean length and tear index.The inner diameter has the greatest influence on the tensile index.The width ratio of bar to groove and bar contact surface have the greatest influence on the tensile index.The adjusted Rsquared of almost regression models are greater than 0.5,which indicates that the regression models have a better fitting accuracy.(3)The bar prediction model PSO-BPNN is constructed and exported in the form of a file for later recall.Based on the construction of the BPNN model,the PSO algorithm is used to optimize the weights and thresholds of the parameters,which eliminated the disadvantages of slow convergence and poor generalization ability of the BPNN model.Its fitting accuracy is evaluated through MRE(Mean Relative Error),training accuracy,verification accuracy,test accuracy,training time and AUC(Area Under Curve),etc.It has proved that the number of hidden layers of the bar prediction model is 1,the number of neurons is 8,the activation function is Tanh,the output layer has 5 variables,its activation function is Softmax,and the calculations number of the prediction model is 85.The test proves that the model fitting precision is good.(4)The design algorithm of bar profile is proposed,and RPP-CAD(Refining Plate Parameterization-Computer Aided Design)system is established.The bar profile design algorithm is proposed,based on the premise that the bar parameter values are output by the PSO-BPNN prediction model,the center angle of the refining plate is restricted to 40° and the position of the longest bar is fixed.Based on typical three-tier architecture model,the RPP-CAD system is established.It is a one-step refining plate R&D system,characterized by simply inputting the increments of pulp performance parameters and refining control conditions,the refining plate digital model is got automatically.The verification and test result shows that the system can accurately predict the bar profile parameters of the refining plate and output the three-dimensional model.The refining intensity model of specific contact surface load which is proposed in the dissertation provides theoretical reference for the refining process analysis,refining plates selection and its optimization design.It's a meaningful attempt to establish the PSO-BPNN prediction model,which can help solve the problem of the complex variable relationships between the bar and refining mechanism and provide a new idea for the subsequent research on the bar.The established bar design algorithm can not only improve the selection and optimization efficiency of refining plates,but also afford a new method for bar design.The constructed RPP-CAD system can provide technical references for the selection and parametric design of refining plates,shorten its research and development cycle,and assist the enterprises to transform and upgrade.