Dynamic Data Modeling And Optimization Of Carbon Fiber Coagulation Process

Carbon fiber composite material is a kind of new material with excellent mechanical properties.It has the characteristics of high strength,large modulus and small density besides high specific strength and high specific modulus.Due to the excellent properties and wide application of carbon fiber,the process of spinning and coagulation of carbon fiber has been paid much attention by domestic and foreign scholars.At present domestic and international research on carbon fiber coagulation process has remained mostly in mechanism modeling.Especially for dynamic changes during coagulation dynamic data modeling is rarely involved.In recent years,with the rise of statistical learning and machine learning,how to use data to establish a more efficient model has become a very necessary research issue.In this paper,the dynamic data of carbon fiber coagulation is based on the mechanism model.The dynamic data model is designed for accurate and stable prediction of the concentration of the fiber during coagulation process,which can help to improve the performance of real-production.The research includes the following main aspects:(1)In this paper,the data features of the parameters in the carbon fiber coagulation process and the correlation of them are analyzed.A multi-kernel SVM data model(MKSVM)is proposed to solve the data feature diversity of the solvent concentration variation in the coagulation bath.Compared with the traditional SVM model with single kernel,the model proposed in this paper is more suitable for the change of solvent concentration in the fiber with diverse data features.The results of simulation and comparison show the accuracy and superiority of the model.(2)The coagulation process of carbon fiber is a dynamic process which changes continuously with time.In this paper,the clustering analysis of the dynamic data in the coagulation process is carried out to train the kernel matrix switching machine(KSM).The kernel matrix can be switched to the best when the window is sliding,according to the data currently loaded into the window.The kernel matrix mentioned is the multi-kernel support vector machine model which is trained with the data of the category center.Then,the sliding window theory is introduced,and a dynamic data model based on sliding window multi-kernel SVM(SWMKSVM)is proposed.The model can effectively achieve the dynamic performance of the model.The experimental results show that the model still maintains a satisfactory model accuracy in the dynamic process of the coagulation process.(3)In order to optimize the effect of the dynamic model in the whole dynamic process,the immune algorithm is used to optimize the parameters of the sliding window multi-kernel SVM,and an immune sliding optimization multi-kernel support vector machine model(ISAMKSVM)is proposed.The optimization mechanism of the immune algorithm ensures that the global optimal solution is found quickly and efficiently.The result is not only the optimum of each sliding anymore,which guarantees the global optimum of the sliding window in the whole sliding process.Experimental results show that the proposed method can find the optimal parameters quickly and efficiently when compared with the traditional particle swarm optimization(PSO)model.The model has obvious advantages in accuracy and stability.At last,the paper summarizes the research content of the whole paper,points out the shortcomings existing in this research,and prospects for further research direction and methods.