Study On The Hardness Of Superhard Materials(Aluminum Alloys,Polycrystallines,Covalent Crystals) Via SVR

As one of the three pillar industries, material is the basic needs for human survival and development. With the rapid development of modern science and technology, how to improve the hardness of related material is put forward higher requirements. Super hardness materials, as a kind of very important material, have excellent performance. In order to improve the hardness of the material for special purposes with good mechanical properties of superhard materials, researchers have carried out extensive and in-depth research on superhard materials. Although some materials with high hardness in mechanical processing industry have been widely used, but there are still many deficiencies which need to be investigated and improved further. For polycrystalline materials, covalent crystal materials and aluminum alloys, different factors will directly affect the hardness of the material. Therefore, how to effectively carry out the analysis of the experimental data, experimental design and optimization, is very important to choose reasonable.In this thesis, the support vector regression(SVR) modeling theory, combined with the particle swarm optimization(PSO) algorithm to optimize the model's parameters, was employed to model/optimize the hardness for polycrystalline materials, aluminum alloy materials and covalent crystals. The main contents are as following:1The SVR was utilized to model/analysis the relationship between 2 parameters(i.e., shear modulus and bulk modulus) and the hardness of 101 polycrystalline samples including diamond-like or rock salt structure. The results show that the mean absolute percentage error(MAPE) for the 18 test samples calculated via the established SVR model reached 4.94%, which is quite smaller than that(27.4%) reported in the literature. The relationship between the ratio k of shear modulus and bulk modulus of diamond-like or rock salt structure materials and their hardness indices are highly complex nonlinear. It was also found that when the shear modulus is 522.627 GPa and the bulk modulus is 1.17693, the maximum hardness value would reach 99.82 GPa.2According to a measured dataset on the elements and related measured hardness indexes of 35 aluminum alloy samples, the PSO-SVR was used to construct the relationship between the composition and hardness. The calculated results were compared with that calculated via a semi-empirical formula reported in literature. The results show that based on the identical 29 training and 6 test samples, the established SVR model has smaller prediction error and higher prediction accuracy and better generalization ability. Factor analysis and optimization is conducted by using the SVR model. It was found that the maximum hardness of aluminum alloy will reach 154.96 GPa while its composition(Si?Fe?Cu?Mn?Mg?Ti?Zn?Cr) is 0.461399%?0.428445%?4.55044%?0.869609%?1.7023%?0.107273%?0.256642%?0.0790808%, respectively, and is greater 3.4% than the maximum hardness found via experiment. The constructed SVR model was also used to analysis the synergistic effect of multi-elements on the hardness of Aluminum alloys.3According to a measured dataset on the structure parameters and related measured hardness indexes of 26 covalent crystal samples, the PSO-SVR was used to construct the relationship between the structure parameters and hardness. In the case of small number of samples, the established SVR model still possess higher accuracy either for 21 training samples or 5 test samples. The MAPE for training samples is 2.5108%, which is superior to that 19.725% calculated by a nonlinear regression model(NMR). The MAPE for test samples is 8.729%, which is better than that 11.277% calculated by the NMR model. The total MAPE for 26 samples is 3.706%, which is smaller than that of 18.101% by the NRM model. The constructed SVR model was used to predict the maximum hardness 137.4GPa of covalent crystal materials under the optimal structure parameters. The results demonstrate that the established SVR model possesses greater predictive ability and generalization ability than the NRM model. Furthermore, the synergistic effect of the structure parameters on the hardness of covalent crystals was also conducted by using the constructed SVR model.The results of this study reveal that the established SVR model can be used to accurately model the relationship of the hardness of the superhard samples. It can provide scientific guidance for the development of novel superhard materials, and would save a lot of human and financial resources and time.