Research On The Prediction Technology Of The Quality Of The Rubber Mixing Based On BP Neural Network

Rubber mixing is an important process of rubber processing.With the progress of technology,the technology of Rubber mixing has also been greatly developed.From the traditional open mixer and the mixer to the present low temperature primary method,continuous mixing,wet mixing and other new technologies,the quality of the Rubber mixing has also been greatly improved with the development of the Rubber mixing technology.However,no matter which kind of refining technology,the homogeneity and stability of the quality of the Rubber mixing can not be well controlled.The main reason is that there are too many factors affecting the quality of the Rubber mixing,which include the raw material performance,process formula,equipment performance,process parameters,etc.,as well as environmental factors and human factors.The factors that have been considered in the existing technology of Rubber mixing quality prediction are all focused on the process formulation and process parameters.The prediction of the quality of the gum making is relatively one-sided.Therefore,it is very important for enterprises to improve the quality of products and reduce the production cost by using artificial intelligence algorithm and advanced computer technology to realize the accurate prediction of the gum refining quantity according to a large number of historical data of the enterprise Righteousness.In view of the above problems,this paper studies the prediction technology of rubber rubber refining quality by using BP neural network and some production data collected from enterprises.Firstly,based on the literature at home and abroad,the paper studies the prediction technology of BP neural network on product quality and the factors influencing rubber rubber quality;establishes the prediction model of rubber rubber refining quality by BP neural network,trains the model;then optimizes the prediction model by using the bequential algorithm and particle swarm optimization algorithm respectively;finally,it uses Matlab platform to compile the prediction model The program software is made and the practical application of the technology of predicting the quality of the Rubber mixing is realized.The main research work is as follows1.Through consulting a large number of relevant literature,the paper studies the prediction technology of rubber refining quality,artificial neural network technology and its application in rubber material field,and makes the research plan and technical route.2.The main technical indexes and influencing factors of the quality of the Rubber mixing are analyzed,and the mechanism of the prediction of the quality of the Rubber mixing is studied.The prediction model of the quality of the Rubber mixing is established by using BP neural network.3.The paper determines 8 variables,such as the ambient temperature,environmental humidity,he refining temperature and the time of the mixer,the distance between the rollers,the number of passes,the time of the glue making,and the time of the glue material,etc.as the input parameters.The Mooney viscosity and tensile strength are converted into the synthetic quality of the Rubber mixing by linear weighting method,and the Mooney viscosity,tensile strength and the comprehensive quality of the Rubber mixing are taken as the output Parameters,a single BP neural network prediction model is constructed.4.In order to improve the accuracy of the above neural network prediction model,the structure and principle of BP neural network are studied.The model is graded from three aspects,such as accuracy,convergence speed and mean square error of the model.The best neural network structure is found and trained for different prediction objectives.The results show that the best training algorithm of the BP prediction model of Mooney viscosity is Bayesian regularization algorithm,the best number of neurons in the hidden layer is 9,the average relative error of prediction is2.24%;the best BP prediction model of tensile strength is momentum gradient descent algorithm,the best hidden layer neuron number is 6,the average relative error of prediction is 4.66%;the synthetic Rubber mixing synthesis The best training algorithm of BP prediction model is Bayesian regularization algorithm,the number of neurons in the best hidden layer is 7,and the average relative error of prediction is 4.8%.5.The GA-BP neural network model was obtained by optimizing the above prediction model by genetic algorithm.The average relative error of Mooney viscosity was 0.92%;the average relative error of predicted tensile strength was 3.34%;the predicted average relative error of synthetic quality of Rubber mixing was 3.4%,which was about 30% higher than BP neural network model.6.The PSO-BP neural network model is obtained by optimizing the above prediction models respectively.The average relative error of Mooney viscosity is2.16%;the average relative error of predicted tensile strength is 4.32%;the predicted average relative error of synthetic quality of refining is 4.4%,which is about 8% higher than BP neural network model.7.Using the method of grey correlation analysis,the influence degree of the input parameters of BP neural network prediction model on Mooney viscosity from high to low is as follows: roll distance of mixer,mixing time of internal mixer,mixing temperature of internal mixer,mixing time of mixer,times of roller passing,stopping time,ambient temperature and humidity;The influence degree of tensile strength from high to low is as follows: mixing time of internal mixer,mixing time of opening mixer,roll distance of opening mixer,mixing temperature of internal mixer,times of roller passing,parking time,ambient humidity and ambient temperature;From high to low,the influence degree on the comprehensive quality of rubber mixing is as follows:rubber mixing temperature of internal mixer,roll distance of open mixer,times of roll passing,rubber mixing time of internal mixer,rubber mixing time of open mixer,parking time,ambient humidity and ambient temperature.