Study On Predictive Model Of Rubber Carbon Dispersity And Its Application

Carbon dispersity of the rubber is an important target to evaluate the quality of mixing rubber and the capability of its products. However the on-line and real-time measure of carbon dispersity based on hardware is restricted in application, so it is an effective method to estimate the carbon dispersity on-line with the mathematic model which is the functional relations between the carbon dispersity and the process variables which is easy to measure. Recently SVM (Support Vector Machine ) is developing effectively and potentially, so it is meaningful theoretical and valuable practical to research the mathematic model based on SVM to predict the mix dispersity on-line.Various influential factors and process parameters about the quality of mixing rubber is deeply analyzed. Then the predictive model of rubber carbon disperisty with the new laboratory mixer based on SVM is established for the first time through the experimental study. Many compounding mixing rubbers could be estimated on-line in this new laboratory mixer. Because some parameters of SVM exert an influence on the predictive result, we optimize these parameters by comparison and analysis. The results of the simulation show that it's an effective method to predict the carbon dispersity on-line by the mathematic model based on standard SVM, and the model's ability of generalization is not bad.In order to predict more accurately, we improve this mathematic model. Firstly, the training samples of SVM are optimized by the fuzzy cluster algorithm according to the peculiarity of these samples. Secondly, the fuzzy system based on integration of SVM is established by learning the optimized samples. The results of simulation show that the improved model achieves high precision and good generalization, and the structure of the improved predictive model is simple, which will take less computation resources of the computer.The monitor control software of the new laboratory mixer based on configuration software is developed. The software has many good characteristics such as friendly interface, reliable and accurate data collection, easy operation etc. And it can predict the carbon dispersity on-line in the process of mixing with the improved predictive...