The Process Modeling Of Tread Production Based On Kernel Method

With the rapid development of Chinese automobile industry, the market demand for tire quality is higher and higher. The tread is one of the most important semi-manufactured goods in tire manufacturing, because its quality determinates the performance of tire-manufactured goods. In this paper the tread extrusion line of actual process and production mechanism of knowledge, the process of production and acquisition of the original sample data for theoretical research and analysis, through the introduction of a new machine learning method--kernel method, establish model on the process. It could be made as the basic of suring optimization of control system designing and prediction.The paper analyzed the prime data from Access database and built up a proper model. The tread extruding line is a complex, nonlinear and dynamic industrial system, and we hadn't known clearly how they work, so it is usually hard to analysis the mechanism of processes, moreover it will take us unbearable staff cost and financial cost. Because of that, we used the support vector machine to analyze the system and build up the model, which is the core representative of kernel method.For the method of the theory in the paper, it introduces the basic theory knowledge of Statistical learning theory and kernel method, especially the basic concepts and principle of Support Vector Machine and regression. On the model selection, it introduces emphatically the kernel function and parameters selection method.Since the acquisition of the original data is contain multiple property relationship of the overall data, the data processing is to temporarily do not need attribute values are removed, and which may influence model to study the error data, pathological data, lag data artificial removal method.In order to simplify the input attribute number, this paper uses PCA (Principal Components Analysis) method of data dimensionality reduction.In the last chapter, introduced the Matlab7.1environment based on RBF radial basis function toolbox profiles as well as the main function, and detailed introduces the system of the basic concepts of modeling, data analysis and modeling process.