Modeling Of Polyester Fiber Polymerization Process Based On Combined Kernel Function

As the largest variety of synthetic fibers,polyester fiber is widely used in all aspects of the national economy because of its good physical and chemical properties.The polyester fiber production process mainly includes polymerization,melt transportation,spinning,and other parts.The polymerization process is the first stage of polyester fiber production,which plays a decisive role in the quality of the final fiber product.The polymerization process has a high degree of nonlinearity and strong coupling.To analyze the polymerization process more deeply,a data-driven model of intrinsic viscosity was established.For the esterification and esterification vapor separation stages,the coupling between multiple variables was considered.A model of oligomer density and oligomer flow was established.The research work of this article includes the following aspects:(1)Analyze the polyester fiber polymerization process,and the production process of the esterification and esterification steam separation stage,extract the key variables of the production process and preprocess the actual industrial data.Aiming at the gross errors in industrial data,the Pauta criterion is applied to eliminate the influence of outliers on the data-driven model,and the normalization and standardization methods are used to achieve the fusion of different dimensional data.(2)Combine the fractal dimension and the partial least squares algorithm to establish the fractal dimension feature selection method based on the correlation dimension,and use the variable projection importance to analyze the selected features.The experimental results show that the fractal dimension feature selection algorithm is better than the traditional feature selection methods.(3)Since STL decomposition can fully extract the periodic and trend characteristics of the time series,the STL decomposition is combined with the kernel function analysis.Based on the Gaussian Process Regression(GPR)model,a combined kernel function based on the STL decomposition is proposed to construct a single-output GPR model.Aiming at the problem of the traditional conjugate gradient method with a large amount of calculation and difficulty in finding the best,the improved conjugate gradient method is used to optimize the parameters of the kernel function.Through the prediction of the intrinsic viscosity in the polymerization process of polyester fiber,the validity of the model is verified.(4)Because of the complex models and the strong coupling of variables in the esterification and esterification steam separation process,a combined kernel function of STL decomposition based on the data weighted fusion algorithm is proposed to construct a multi-output GPR model(MGPR).Aiming at the problem of a large number of parameters in the multi-output model of the combined kernel function,the immune particle swarm optimization algorithm is used to optimize the model parameters,which improves the generalization ability of the algorithm.Through the prediction of the density and flow rate of the oligomer generated in the esterification tank in the esterification and esterification steam separation stage,the effectiveness of the MGPR model is verified.