The Effect Study Of Preparation And Tribological Properties About PI Based On Support Vector Machines

Polyimide(PI) is a special engineering plastics, one of the best overall performance polymer materials. The excellent tribological properties have expansive applications in the field of tribology. Experimental study about the preparation process of polyimide existences the time-consuming problem. While the establishment of mathematical model studies the relationship between imidization process about the PI and friction and wear properties. For reducing the amount of experiments, cost savings, improving the efficiency of research, which has important significance.In this paper, prediction model based on the least squares support vector machine(LS-SVM) established PI thermal molding process conditions and their strength. Heat molding conditions selected for the curing time, molding pressure, molding temperature.strength conditions for flexural strength and tensile strength.the sample data were obtained using orthogonal experiment plan arrangement. Using LS-SVM prediction model and compared with the BP neural network model of prediction results to prove its feasibility. Research the relationship between heat molding conditions and the strength of PI, got the best heat molding conditions. Then preparation of PI specimen preparation with optimal thermal molding conditions to carry on friction and wear test, explored the relationship between preparation conditions and friction and wear properties of PI, the LS-SVM was also used to establish model of the preparation conditions, friction and wear properties of PI. Selection of the imidization conditions for the amount of catalyst,dehydrating agent and imidization time, friction and wear properties of the wear rate and friction coefficient, the sample data obtained by testing, the LS-SVM prediction model was established, compared with BP network model, Research the relationship between PI preparation conditions and the friction and wear properties, and got better performance PI samples. The main contents were as follows:According to the PI heat molded test sample data of the tensile strength and flexural strength, optimizing LS-SVM model of penalty coefficient and nuclear parameters under MATLAB software, the hot-molded LS-SVM prediction model was established, and thehot-molded BP neural network prediction model was also established, the best prediction model was achieved through comparative analysis and eventual experimental verification.Based on friction and wear test sample data, optimizing the penalty coefficient and kernel parameter of LS-SVM model, the LS-SVM prediction model of friction and wear was established, with the BP neural network model and genetic algorithm optimization BP neural network model comparison study, which was verified feasibility for forecasting the tribological properties of PI.Finally, studied the influence of PI imidization conditions on the friction coefficient and wear rate. this paper focuses on friction and wear properties of PI, compared by BP neural network and LS-SVM, two ways optimizing the friction and wear properties laid a foundation for the future optimize performance prediction method of PI and other materials.