Study On The Calculation Of Macroscopic Properties For FCC Metals Based On Molecular Dynamics Simulation And Artificial Neural Networks

With the rapid development on computer simulation technique, computer simulationhas become an important tool for studying micro and macro behavior on material science.Molecular dynamics simulation and artificial neural networks are two importantsimulation methods . The former is used to study the microstructures of system, and thelatter is used to study materials designing and properties predicting. In this thesis, based on the basic principle of the computer simulation techniques andthe method combined molecular dynamics calculation and neural networks, the softwarepackage was designed by ourselves. By adopting the software package, the elasticproperties such as the bulk modulus, shear modulus, and elastic constants, thethermodynamic properties and the surface energy of some fcc metals. Especially thehigh-index surface energy of fcc metals and alloy were calculated by the software packagesuccessfully. Using the molecular dynamics simulation software package, the elastic properties(such as the bulk modulus, shear modulus, and elastic constants) were calculated , theresults are in good agreement with experiments available. By using the EAM potentialsproposed by Johnson and the lattice dynamics method, the thermodynamic properties(such as the phonon spectra, state density, specific heat, et al.) of fcc metals werecalculated, and compared with experiment data and results calculated by using othermodels. It's found that the results for fcc metals are successful. Via molecular dynamicssimulation, surface energies of three low-index and some high-index planes have beencalculated. Combined the data calculated with the molecular dynamics model, a great dealof data were trained many times and compared with the calculated data, the BP modelpredicting high-index surface energy. The results show that the method has high predictingaccuracy. The order of the three low-index planes was predicted exactly. The surfaceenergies on the other planes show a tendency that first increasing and then decreasing withincreasing angle between the planes and (111) plane. When the angle comes to 40°, thesurface energies reach the maximum. And the average values of surface energy are closedto the experiment data. Since the experiment data are the average values of all planessurface energy and surface energy of the high-index planes are higher than the low-indexplanes, so the calculated results are lower. And the results show that the surface energy forNi3Al alloys may not be represented as averaged concentration values of thecorresponding surface energy of the pure components. The surface energy value is closerto that of pure Ni. The reasonability and accuracy of the model can by checked by calculating themacroscopic properties and comparing it with experiment results, which is very importantfor the development of the calculated model.