Abnormal Analysis And Prediction Modeling Of The Quality Of Tissue Paper

In order to improve the quality of the product.Aiming at the loopholes and quality fluctuating problems in the quality test of the tissue paper before processing in the actual production,we collected the quality data of all the qualified tissue papers in the same time period by investigating A company.There were 563 sets of data.Based on these data,multivariate statistical process control(MSPC)is applied to the quality control of tissue paper,and the contribution rate of comprehensive variables is calculated to analyze the cause of the anomaly.Finally,through the introduction of 10 sets of fault points to verify the accuracy of the model,all the results are reported anomaly,indicating that the research method for quality monitoring is effective and feasible.Of the 563 groups of products qualified for quality inspection,65 groups anomalies are obtained through MSPC model monitoring,of which the 7 groups is of a high degree of abnormality.The abnormal detection rate in qualified products is up to 11.55%.This model can greatly improve the monitoring of the product and improve the overall quality of the product.According to the practical production,there are some key process variables that effect product quality can not be measured online,because the tissue paper pulp fiber morphology important influence on the quality of the paper,In order to establish a soft sensor model of fiber morphology,implementation of advanced control in the process of refining.In the A enterprise,we collected the fiber morphology of the lap pulp,fiber morphology under different pulping time and the fiber morphology under different refining conditions and the parameters of the production process.The model is trained by the collection of data sets,and the soft sensor model of fiber morphology is built based on support vector machine regression(SVMR).Validated by the collected data,the MRE(mean relative error)of the fibers weighted in length is 2.84%,rate in length of MacroFibrills was 5.65%,the kinked fiber percentage is 3.12%,the fibers width is 2.91%,the fibers coarseness is 3.09%,the percentage of fines(%in length)is 2.91%,the broken ends fiber percentage is 3.89%.The model errors are in a good range,indicating that it can improve the quality of pulp and adopt low energy consumption,low cost and high production process..Finally,aiming at the problem of the large fluctuation of tissue paper quality,the waste of raw material and the equipment is often running at high energy consumption position,in order to achieve the online prediction of product quality,we should control the production process under the guarantee of product quality.In this paper,the model is trained by the data collected on the workshop.Through the exploration of three algorithms,LASSOCV,SVMR and GBRT,it is found that the precision of the GBRT(gradient boosting regression tree)algorithm is the best.It is verified that the geometric mean value of tensile strength 's MRE is 7.83%,the geometric mean value of softness is 8.06%,and the bulk is 4.47%.The results show that the error is controlled in a good range,which can realize the prediction of the quality of tissue paper,reduce the fluctuation of product quality,reduce the cost of raw materials,and achieve energy saving and production increase.