Support vector machine (SVM) is one of the most widely used algorithms in parameter inversion of rough surface. However, the penalty parameters (C) and kernel function parameters (G) in SVM affects accuracy of results. If the parameters are not used properly, the model will lead to "over learning"or "less learning", which reduces greatly prediction accuracy. Several optimization algorithms for C and G of SVM are shown, such as K-fold cross validation (K-CV), genetic algorithm (GA) and particle swarm optimization (PSO). An improved PSO algorithm based on K-CV and GA (GA-CV-PSO) is proposed. The training set and test set are constructed with rough surface backscattering coefficient obtained by the moment method (MoM). Inversion precision and calculation time of different optimization algorithms are compared. It shows that GA-CV-PSO algorithm overcomes shortcomings of single optimization algorithms, with more accurate inversion precision and stronger generalization ability.

Non-Fourier heat transfer model cannot be solved by conventional method of separating of variables. A conventional method of separating variables is developed. Fourier model, Cattaneo-Vernotte (C-V) model and dual phase lag (DPL) model are solved by the method. Differences of temperature fields calculated with three models are compared. Temperature fluctuation in C-V model and DPL model is studied. And relation between temperature fluctuation velocity and time lag of heat flux is studied.

Reduced-order model of proper orthogonal decomposition method (POD) was established to analyzed performance of fluid flow and heat transfer in flat tube bank fin heat exchanger. Numerical results of POD were compared with those of FVM. It indicates that average relative error between POD results and FVM results is no more than 5.7% while computational speed is improved 164.78-2021 times. Conclusion is of theoretical meaning for optimizing heat exchanger structure and running mode in engineering.

Current study shows that range distribution of ions implantated into plant seeds obtained with TRIM program are far from experimental data. According to characters of microstructure of plant seeds, microstructure models and calculation programs are designed. With Monte-Carlo simulation range distributions are calculated for iron ions implanted into peanut, color cotton and wheat seed at different energies (110 keV, 20 keV, 200 keV) and different doses (2×10¹⁶ ions·cm^-2, 5×10¹⁶ ions·cm^-2, 10¹⁷ ions·cm^-2, 2×10¹⁷ ions·cm^-2). The results agree well with experimental results. A random sampling simulation method is obtained for interaction between implantation ions and seed microstructure, which provides an idea for theoretical study of interaction between implantation ions and organisms.

We obtained a class of solitary wave solutions of modified Benjamin-Bona-Mahony (mBBM) equation with kink-antikink structure by using hybolic-function expansion method. Solitary wave solution reduces to a kink-like solution or bell-like solution under different limitations. We analyzed structures of solitary wave with double kinks. Dynamical stability is investigated numerically with a finite difference scheme. The scheme is implicit and it is absolutely stable in linearization sense. It indicates that single soliton with double kinks is stable under different disturbances. Meanwhile,collision of two solitary waves is numerically simulated. It was found that collision between two solitary waves can be either elastic or inelastic.

We study carbonation process of CaO with lattice Boltzmann method. Specifically,effects of porosities of CaO particles on carbonation rate are investigated. It shows that efficiency depends on porosity nonlinearly due to competition between molar mass and permeability of particles.

Smoothed particle hydrodynamics (SPH) is applied to simulate parametric sloshing in various shape aqueducts. Taking rectangular tank as an example, the first symmetric mode of sloshing were investigated, including free surface elevations and frequency relationship, which confirm characteristic of nonlinear parametric sloshing. And the first two anti-symmetric and symmetric mode of parametric sloshing for U-shaped and circular aqueducts were also studied and presented in detail. These results confirm validity of SPH method for simulations of parametrically excited sloshing motion.

A detailed comparative study is made to investigate three boundary treatment schemes in lattice Bohzmann methods (LBM).They are half-way bounce back scheme,sub-grid scheme,and a scheme to treat curved boundaries.It is found that particle translational and angular velocities calculated with half-way bounce back scheme oscillate strongly,while particle velocities obtained by sub-grid scheme evolve smoothly.

An explicit equation for critical quantity of heat in condensing flow in a shock tube was derived.The equation is of general validity and applies tO vapor flow with or without cagier gas.It shows that critical quantity of heat in internal heat flow for phase change is greater than external heat flow.As flow is choked thermally,a shock wave is generated and propagates upstream.A location where local static pressure begins to rises,which has long been regarded as onset of unsteady condensation in a steady supersonic nozzle,is not acceptable in shock tube flow.It is before the zone of condensation.Furthermore,a second-order AUSM scheme is applied tO two-dimensional compressible flow with condensation.Computation results reproduce integral predictions well.

With molecular dynamics study,we investigate microstructure of liquid Al₂O₃ at 2663 K and 2223 K.The first and second nearest-neighbor distances are in good agreement with results of Ansell et al.Analysis on coordination numbers and angle-distribution functions show that the microstructure of liquid aluminum oxide mostly consists of three-fold coordination of oxygen,four-fold and five-fold coordination of aluminum.Evidence of liquid-liquid phase transition due to separation of high and low coordinate species is not found.

Generalized coarse-mesh finite difference(GCMFD) method,which uses generalized-geometry coarse meshes to speed up the method of characteristics(MOC),is proposed.A method adjusting width factor automatically is adopted in the GCMFD method.The GCMFD method is adopted in a 3-D neutron transport MOC code TCM.Numerical tests are used to verify acceleration effect of GCMFD in TCM.It shows that the GCMFD method can use generalized-geometry coarse meshes to accelerate MOC iteration,and the GCMFD method with automatically adjusting width factor has good acceleration effect.

Kelvin-Helmholtz instability is carried out with local Steger-Warming flux splitting method and NND scheme for hydrodynamic equations. Linear growth rates agree well with linear stability analysis. The method provides clear interface deformation images.

In one- and two- dimensional approximations,the penetration depth-concentration distribution for 200 keV vanadium ions implanted into peanuts is simulated using a Monte Carlo method.The calculation is in good agreement with experimental results.The depth-concentration distribution for nitrogen ions with low energy implanted into peanuts is calculated which can not be obtained with experiment.It provides a computational method for the depth-concentration distribution of ions with low energy implanted into seeds.

According to the characteristics of perfectly matched layers (PML), a convolutional PML (CPML) is chosen to truncate open boundaries used in the numerical simulation of low-altitude electromagnetic pulse. On the basis of the split-field PML and the plane-wave solution of electromagnetic field in a free space, an unsplit-field PML is constructed. With convolutional theorem in the Fourier transformation, discerete iterative equations of electromagnetic field components in the CPML media are presented in three-dimensional prolate-spheroidal coordinate system. The effectiveness of the method in numerical simulation of nuclear electromagnetic pulse is shown.

A three-dimensional numerical simulation of the temperature field in high temperature combustion furnace is performed using a method of discrete ordinate to solve the radiation transfer and transfer coupled equations.A computer program of flow, combustion, heat transfer and NO_x turbulent formation is developed. The temperature distribution of Fuel Direct Injection (FDI) is numerically analyzed at different preheated air temperature and the validity of the applied numerical simulation is tested. The simulation results show that when preheated air temperature increases the maximum temperature gets larger, the temperature gradient becomes lower and the flame length longer.They are in agreement with experiments.

The three-dimensional numerical simulation of the non-premixed combustion turbulent mixing processing in high temperature combustion furnace have been performed with authors developed computer program of flow,combustion,heat transfer and NO_x turbulent formation.The 3d-distributions of mixture fraction and its turbulent fluctuation are forecasteed under three different inlet air temperature conditions in combustion furnace.Compared with non-premixed combustion,the results demonstrate that,under certain geometry and dynamical conditions,the turbulent mixing of high temperature air combustion proceeds with smaller intensity within a larger zone; the turbulent mixture fluctuation distributes within the larger combustion zone.This also indicates that the flame thickness appears to be larger and that heat is released within larger combustion zone.The simulating conclusions are in agreement with the results of similar experiments.

Using the Modified Analytic Embedded Atom Method(MAEAM),the phonon spectra and lattice specific heat of bcc metals(Cr,Fe,W,Mo,Ta,V,Nb) are calculated.The obtained results are in good agreement with the experimental data.It is improved greatly as compared with the calculated results by the Johnson model.And the influence on EAM model of the characteristics of bonding orientation of these transition bcc metals is discussed.

Mold-filling of polymer melt in three dimensional thin parts is a viscous incompressible flow with moving boundaries.Based on generalized Hele-Shaw flow theory,the finite element systems for pressure fields are constructed by the Control Volume Methods.The advancing of melt front is traced by a hybrid FEM/Control-Volume method and the dynamic simulation of the mold-filling process is achieved.

Based on the dual mixed variational formulation for the Signorini problem,a nonconforming finite element method is proposed.The discrete B-B condition is confirmed and the error estimation O(h^3/4)for Raviart-Thomas(k=0)finite element is achieved.A Uzawa type algorithm is used for solving the Signorini problem which is discretized by conforming finite element method as well as nonconforming finite element method.The accuracy and efficiency of both are demonstrated by numerical results.By contrast to the conforming one,nonconforming method is more cost-effective.

The Rayleigh-Taylor instability is numerically studied through the Large-Eddy-Simulation (LES) approach.The whole evolution process of the instability is achieved,including the linear stage,the regular nonlinear stage,the irregular nonlinear stage and the turbulent mixing stage.The characteristic of the interfacial motion and the growth of the mixing layer thickness are analyzed.And the averaged turbulent energy as well as the flux of passive scalar is calculated at both the resolved scale and the subgrid scale.The LES method is proved to be an effective approach for the simulation of the Rayleigh-Taylor instability.

The Euler/Kirchhoff method is adopted to calculate the far field noise of UH-1H rotor,and the calculated results are compared with the measured data.The effect of rotor blade tip shape on high speed rotor noise is investigated.The results show that,with the decreasing of airfoil thickness near the blade tip or the increasing of the taper ratio near the blade tip and the sweep of the blade tip, the high speed rotor noise can be reduced to some extent.

The laminar flow and heat transfer of water in periodic fully-developed divergent-convergent channels has been investigated numerically.The SIMPLE algorithm and body-fitted coordinates (BFC) are used.The periodic boundary condition is treated using the method proposed by Amano.It is found that under the range of Re=100~1 000,the friction factor is about 10%~200% more than that of parallel plate channel,while the average Nusselt number is about 40%~320% more than that of parallel plate channel.

A finite element method is used to numerically study the steady axisymmetric incompressible flow between two concentric rotating spheres. The non-uniqueness of the supercritical flow exhibits three different modes with zero,one and two Taylor vortices at each hemisphere.

The description of the moving grid-strategy is presented. Based on transfinite interpolation, an O-H type grid around one blade is generated. In each subdomain, the weighting funtion is given to account for the time-variation of its nodes. The multi-block grid modelling the whole rotor is achieved by the construction of grid-blocks around each blade which are connected with coincident nodes on their common boundaries, the flowfield of a rotor is calculated by considering the transport of interface information. The calculated results show that the moving grid possesses feasibility and effectiveness.

Differential algebra is a new mathematical method and a promising tool in charged particle optics. As a kind of automatic differentiation techniques, it is based on the nonstandard analysis and normal power series theory. It presents a straightforward method for the study of arbitrary order property of nonlinear systems, with accuracy limited only by the computer. The differential algebraic method is widely applicable, and it is an effective and elegant tool for the tracking, analysis and correction of high-order aberrations, sensitivity analysis and optimization of structure design, etc., of the charged particle optical systems. The principle and virtues of differential algebra are summarized. The application of differential algebra method in charged particle optics and its implementation in various computer languages are analysed. Some new research subjects in this field are discussed.

The System generated electromagnetic pulse (SGEMP) generated by X rays in various ranges of fluence have been numerically studied.Certain trends about the electric and magnetic field and their distribution as a function of fluence are abtained and discussed.

A numerical method is given to solve the relavtivistic equations with source terms by finite difference method, QGP phase transition taken into account.The numerical experiments verifies the effectiveness of the method.

A method of couple node velocity is given for reducing distortion of quadrilateral mesh in Lagrangian numerical computation of two dimensional compressible hydrodynamics. It is verified that this method is equivalent to introducing four order smooth term with dissipation property in the momentum equation, thus it plays a role of damping the unphysical oscillation of velocity.

A new method is presented for calculating the quasi steady transonic flow over a lifting rotor blade in hover by using Euler equations. The approach is to solve the Euler equations in a rotor fixed frame of reference using a finite volume method. The results are verified by comparison with wind tunnel data. In cases considered, good agreement is found with available experimental data.

EPM method for two dimensional elastic-plastic hydrodynamic calculation, which couples Eulerian and Lagrangian methods, is presented in this paper. The difference scheme is constructed by means of contour integral. This method includes sliding interfaces treatment, several generating meshes technique, and so on. It can be widely applied to calculations of many hydrodynamic problems.