SVM-based Composite Material Properties Of CFRW Analysis And Prediction

Carbon fiber reinforced wood (CFRW) composites are generated as a new kind of functional material with the improvement of science and technology. Study on this material has made some progress, but the study on its performance and characteristics is still less. Its different performance and characteristics makes different application range and environment, so the properties of CFRW composites are necessary to study. Use of CFRW composites conductivity not only meets the conductive requirement instead of metal materials in the corrosive environment, but also meets the needs of a number of other special applications, therefore, the study on carbon fiber reinforced wood composites conductivity has gradually become one of its main directions.In this paper, carbon fiber reinforced wood composites were preparated through hot-pressing mold technology. The mechanical properties were measured in experimental measurement of CFRW composites with different carbon-fiber ratio, such as static bending strength, elastic modulus, thickness swelling, internal bond strength and so on, but also the surface resistivity were measured in different temperature. On the basis of the data analysis of the relation on the static bending strength, elastic modulus, thickness swelling, internal bond strength, surface resistivity and carbon fiber ratio, and the relation on surface resistivity and temperature, conductive properties influencing factors were determined as carbon fiber content, wood fiber content, elastic modulus, temperature, production process means.Support vector regression machine is a kind of special small sample learning machine based on statistical learning theory. It has strong generalization ability due to the structural risk minimization principles, so as to overcome the over-learning shortcomings of neural networks. In this paper, the conductivity forecast model of CFRW composites was established based on support vector regression machine to aim at nonlinear characteristics of electrical conductivity. There were many influencing factors for the conductivity of CFRW, so a five-input and one-output conductivity model was established to simulate and forecast its performance, with carbon fiber content, wood fiber content, elastic modulus, temperature, and production process regarded as the main influencing factors. The impact on forecast results caused by different support vector machine parameters was analyzed. The results show that the model has high forecast accuracy and strong generalization ability, using the grid search method to select the optimal parameters, theε-SVM andν-SVM modeling to forecast resistance ratio is effective. This study is significance of guiding CFRW functional materials design and research and has certain industrial practical value.