| Nowadays, as the natural mineral resources become less and less, and it is more and more difficult to exploit. This has become a trend that the increasing demand forced us to synthesis new materials artificially so as to replace the natural products. It is well known new-type carbon materials have played an important role in artificial synthetic material. Owing to its excellent electrical properties, mechanical properties and thermal properties, it is widely employed in various fields. Although various types of carbon-based materials have been studied and widely used, the synthesis mechanisms are not all perfectly clear. In this case, to establish a stable and accurate prediction model has a great help for guiding the synthesis process.This thesis aims at the modeling and optimization on the diameter and yield of single wall carbon nanotubes(SWCNTs), the hardness/crystallinity of diamond-like carbon films/diamond film deposited via Chemical Vapor Deposition(CVD) approach.The research process is as follows:(1) Collect the relevant experimental data and modeling.(2) Test the accuracy of the established model.(3) Analysis the related rules revealed by the constructed model.In this thesis, the particle swarm optimization(PSO) method combined with support vector regression theory were utilized to conduct the modeling. The main work includes the following contents:(1) The relationship between the process and diameter of SWCNTs synthesized by gas arc-discharge methodIt was reported that the diameter of SWCNTs fabricated via gas arc-discharge method is mainly affected by several experimental parameters including type of metal catalysts, kind and total pressure of gas and discharge current. Based on the related experimental dataset, we implement the modeling while the above factors as the independent variables and the diameter of SWCNTs as the output. The validation samples were adopted to further verify the accuracy of the established SVR model. The result revealed that the predicted values by SVR model are in quite good agreement with the experimental results, the mean absolute percentage error(MAPE=3.98%)achieved by SVR model is superior to those(17.12% and 14.9%) of artificial neural network(ANN) and multiple linear regression method(MNR) reported in the literature,respectively. At the same time, the established SVR model was used to calculate the sensitivity of each process parameter to the diameter of SWCNTs. It was found that thetwo most sensitive factors influencing on the diameter are the pressure and the current.(2) The process and yield optimization for synthesis of carbon nanotubes by methane decompositionThe mean absolute percentage error 1.85% calculated via the established SVR model for ten test samples is smaller than that(4.36%) of multivariate nonlinear regression(MNR) model. The deduced maximum CNT yield 643.29% is available at a reaction temperature 759.484℃, a methane partial pressure 0.796405 atm and a catalyst(Co–Mo/Al2O3) weight of 0.383613 g, which was calculated by the constructed SVR model and is greater than that(607.00%) predicted by using the MNR model. The interactive effects of three process factors on the yield of CNTs are graphed via grid-screening. It is concluded that the PSO-SVR methodology can establish a stable and reliable theoretical model for synthesis of CNTs by methane decomposition, which can provide a valuable guidance for searching of the optimal process condition to get maximum yield.(3) The process optimization for depositing diamond film via microwave plasma enhanced CVD methodThe smaller of the half high width(FWHM) of the Raman spectra of Diamond crystal orientation(111), the higher of its crystallinity. In the experiment of microwave enhanced chemical vapor deposition of diamond films, we conduct the modeling,testing and process analysis based on a relevant experimental dataset while as the diamond powder particle size, substrate bias voltage, the concentration of methane,hydrogen flow rate, microwave power, chamber pressure, bias time, deposition time as independent variables and FWHM as the dependent variable. It was predicted by the established SVR model that the smallest FWHM may be 3.85242cm-1 when adopting the optimal process combination(diamond powder particle size 6165.63 mesh, substrate bias voltage 18.8991 v, the concentration of methane 0.46851%, hydrogen flow rate241.172 sccm, microwave power 1045.55 w, chamber pressure 41.9995 Torr, bias time20.3574 min, deposition time 9.59786h). In addition, the grid-screening intuitively reveals the reaction rules about the effects of preparation process on the crystalline of diamond films.(4) The process and microhardness optimization for diamond-like carbon film fabricated by plasma enhanced CVD depositionIt was reported that the microhardness of diamond-like carbon film deposited via microwave enhanced CVD is mainly affected by the argon concentration, chamber pressure and RF power and substrate bias factors. Based on the related experimentaldata, the argon concentration, chamber pressure, RF power and substrate bias were taken as the independent variables, the PSO-SVR was used to model the microhardness of the diamond-like carbon films. The stability and accuracy of the established SVR were analyzed and compared with those of the MLR model. It is evaluated that the SVR model achieved average absolute error of 0.22 nm, RMSE 0.48 nm and MAPE 2.7%, all of them are better than those(0.49 nm, 0.59 nm, 5.83%) achieved by MLR model. The influences of independent variables on the dependent variable was displayed by using3 D image, which is consistent with the experimental result, which proved the accuracy of the established SVR model from another perspective. Finally, the optimum process parameters were also obtained by using the SVR model.It is concluded that the SVR can accurately, effectively applied to model for the SWCNTs diameter, the hardness of diamond-like carbon film/diamond film. The constructed SVR models are robust, accurate, and can quantitatively reveal the effect of various experimental parameters on the dependent variable, which has the important theory significance and practical application value to save time, reduce the experiment cost and the number of tests, etc. |
PDF Full Text Request