Study Oil Extrusion Temperature Prediction Model Of Tread Triplex Extruder Lines

As the rapid development of the automobile market,more and more kinds of tire areproduced, the requirement of tires on wearabilitiy, good surface-appearance and highmechanical perfomance are also improved at the same time, all these generate a great progresson the tread composite extrusion technologies. Though the current tread triplex extrusionlinkage lines have many advantages on processing technology, there are still some kinds ofdisadvantages like the inconsistent of extrusion temperature and pressure result from themachine’s complicated structure. The research reveals that the tread triplex extrusiontemperature is a composite indicator involving many factors durig the extrusion, so it’s ofgreat significance to improve the production capacity and extrusion quality of thesemi-finished product if we can control the extrusion temperature. However, most of tirecompanies still measure the extrusion temperature by manual opration or laser thermometers,due to kinds of subjective factors, the manual measurement method often brings undesirablephenomenas like inaccurate result,low measurement speed, feedback lagnegative,moneyconsuming and so on, while instrument measurement methods is vulnerable to workingenvironment and proximity produciton lines, online measurement accuracy is also not meetthe production needs, the soft sensor is a better way to solve the above problems.Based on the principle of auxiliary variables’ selection of the soft sensor method, thisdissertation firstly made a in-depth analysis to the basic structure, working principle and theextrusion process technologies of the triplex tread extrusion lines, then discussed theinfluencing factors on tread extrusion temperature, which provides a reasonable basis for thechoice of auxiliary variables. On this basis, each of the three extruders’24process parameterssuch as screw speed, pressure of the tube nose, the temperatures of each section, the nosetemperature of composite mold, the thickness of mouth-like plate and so on are choosed as theauxiliary variables, and a further step to filter the variables are taken using the principalcomponent analysis. After the analysis, the multiple correlation between the variables arefound, considering the strong abilitiy to simplify the multi-variable data table and the widelyapplication in the industrial field for its high efficient in variable dimensionality reductionfunction of the principal component analysis, we establish the tread triplex extrusiontemperature prediction model by principal component regression method firstly. It was foundthat although the principal component regression could overcome the multiple correlationbetween the variables, but the symbol of the partial regression coefficients were still notconsistent with the traditional experience of the extrusion process, making the interpretation of these regression coefficients difficult. Analysis showed that the reason the problem was thatthe correlation between the ingredients and dominant variables had not been fully taken intoaccount. However, the partial least squares method in the extraction of ingredients both takeinto account the ingredients’ greatest containing of original variables’ information and themaximum of the correlation between the ingredients and dominant variables, which combinesthe advantages of a variety of methods such as the principal component analysis, canonicalcorrelation analysis and multiple linear regression, having superior soft sensor modelingcapabilities. Therefore, further steps have been taking to establish the extrusion temperatureprediction model by using partial least squares regression method, model-aided analysis andforecasting results demonstrate, to obtain a better prediction, but due to the variables’ differentsampling positions, sampling lags emerged during the sampling process, which is easy toaffect the precision of the model’s prediction by bring in the gross error into the sample space.Considering this point, this study uses a method of combining the mean filter and3σ criteriaon the sample data correction processing to overcome the adverse effects brough in by thesampling lag, so that the prediction accuracy has been significantly improved and largelymeet the needs of practical production.In addition, since the current mainstream statistical analysis software SAS/SPSS on partialleast squares regression algorithms have been modular, especially the related SPSS modulesare commercially available, in order to improve the economic practicality of the soft sensor intire manufacturing, the PLS algorithm program in the MATLAB compiler environment arewrote and the analysis result are compared with the SAS modular’ calculating, which showsthat the two results are basically the same. These laid the foundation for the furtherdevelopment of intelligent software using in rubber production area and its efficientpackaging. Finally, a preliminary study is carried out on the application of the extrusiontemperature soft sensor prediction model and the intelligent control on linage extrusion lines.In summary, the subject studied is much more complicated than the past, and the researchmethods have a certain degree of innovation. The research not only bulid a theoretical basefor optimizing the rubber extrusion process parameters, but also made a great contribution forimproving the control quality in tread triplex extruder lines, which is in favor of a goodquality and efficieny on tread extrusion.