To study performance of graphic processing unit (GPU) for computational fluid dynamics, simulation on interactions between shock waves and a flame interface, a typical compressible reactive flow, was carried out on CPU/GPU heterogeneous system. Several optimal strategies are taken to raise GPU code performance. Computational results and acceleration performance of GPU with different grid number were examined. It was found that computational results by parallel GPU are the same as those by traditional CPU based on MPI parallel of 8 threads. Computational times of two parallel methods linearly grow with increase of computational grid numbers. Compuational cost by GPU is less than cost by MPI. As grid number is small(1.6×10⁴), speedup ratio of 8. 6 was achieved on GPU. As grid number grows, speedup ratio decreases. However, a ratio of 5.9 still can be held on GPU when grid number is more (4.2×10⁶). Arithmetic on CPU/GPU heterogeneous system provides a good solution for large-scale computations of compressible reactive flows.

To deeply understand mechanism of shock wave interacting with a flame,a numerical study on a planar incident shock wave and reflected wave interacting with a spherical flame was carried out with two-dimensional Navier-Stokes equations coupled with chemical reaction. Okubo-Weiss function are proved appropriate for compressible flow by analyzing characteristic equation of velocity gradient tensor. Emphasis was placed on two-dimensional flow topology characteristics of flame zone. It shows that integral of Okubo-Weiss function conserves within flame zone after passages of shock waves. However,inner and surface of flame zone display completely different flow status. Flow compressibility basically has no effect on evolution of flame. Besides,flow topology of flame zone is mainly controlled by foci and saddle,which means deformation is dominant in flow field.

Based on interconnection transmission differential equation,transform voltage and current expressions are derived by incorporating appropriate boundary conditions for interconnect energy analysis.We considered a practical transmission line with driver and load to find out relation between transform input and output voltage and current responses.Interconnection energy distribution is obtained,which considers non-ideal step stimulation.In 65 nm CMOS process,the non-uniform interconnect analytic model enables to estimate energy within 5% compared with Hspice simulations.The method has high accuracy.The analytic model can be applied to front end optimizing design and analysis of high performance global interconnection.

A three-dimensional analytical heat transfer model for stacked chips is developed, which takes into account horizontal heat transfer effect in three-dimensional integrated circuits (3D ICs) with through silicon via (TSV). Effects of horizontal heat transfer is analyzed with number of strata, TSV density, TSV diameter and thickness of BEOL layer under specific process and thermal parameters. It indicated that temperature rise simulated by the model is lower compared with result not considering horizontal heat transfer effect. Difference of temperature rise can be above 10%. Effect of horizontal heat transfer on thermal management of 3D ICs is more obvious with increasing of integrated level. The model conforms to actual situation. It is more accurate in analyzing temperatures of stacked chips in 3D ICs.

With consideration of via effect and heat fringing effect, a thermoelectric simulation method is proposed which modifies node heat flow due to temperature distribution. Based on thermoelectric duality, thermal resistance models including inner/inter-layer and vias are presented. Take advantage of feedback relationship between heat and electric, the node network heat flow model is modified with temperature distribution. Multilevel interconnects temperature distribution with polymer and silicon oxide as insulator dielectric are analyzed. Compared with results of finite element, the relative standards deviation of the proposed method can be reduced by 71.2% and 12. 9% respectively than those of available models. With consideration of via effect and heat fringing effect, we calculate peak temperature rise in different technology nodes. It shows that interconnect temperature distribution is overestimated in traditional models.

A three-dimensional simulation on deformation and instability of low-density spherical bubble induced by incident and reflected shock waves is performed with Navier-Stokes equations.A computational model validated by experiments is used to study deformation of bubbles,formation and three-dimensional instability of vortex rings.It is found that low.density spherical bubbles can deform and form vortex rings with different rotating directions along streamwise direction induced by incident and reflected shock waves. Baroclinic effect is the main reason for formation of vortex rings.Vortex rings move due to self-induced effect and flow field velocity. And azimuthal instability OCCURS due to disturbance.Finally.vortex rings may induce turbulent flow a8 they form a complicated structure dominated by small-scale streamwise vortices.

With influences of interconnect inductance,thermal-electric coupling effects and interconnect temperature distribution,a delay optimized method by repeater insertion is presented.A interconnect resistance model and a delay model are obtained respectively based on interconnect temperature distribution.The repeater insertion optimal delay is calculated considering electro-thermal coupling among power,delay and temperature.Optimized results are successfully obtained with Matlab software.Repeater insertion in 45 nm technology is simulated.It shows effectiveness of the method.In addition,it indicates that the optimal delay is overestimated without inductance effect.Optimal delay is underestimated without consideration of temperature distribution.As overestimated global interconnect width is 245nm,8.71% optimal delay can be underestimated without considering temperature distribution effect.

A variable-density stochastic vortex model is used in simulating H₂/O₂/N₂ turbulent jet diffusion flames,in which turbulence is modeled by vortex sample,vortex suppression and vortex turnover.A large-vortex suppression mechanism for variable density reacting flows is presented and effects of parameters in model prediction are discussed in detail.It shows that the modified model can predict reasonably H₂/O₂/N₂ jet flame structures,and reflect characteristics of turbulent vortices.It also reveals that various parameters in vortex sample and vortex suppression affect results of the model.

A joint-scalar probability density function (PDF) method, a steady laminar flamelet model, a Eulerian unsteady laminar flamelet model and a hybrid RANS and PDF method are employed in calculating a bluff-body stabilized turbulent non-premixed Sydney flame HM1. Nitrogen oxides emission is numerically simulated with the joint-scalar PDF method and the Eulerian unsteady laminar flamelet model in order to investigate influence of combustion models on the formation of nitrogen oxides. Comparisons between numerical results and experimental data show that chemical reactions by PDF are best while the computational cost is expensive. The laminar flamelet model results shows lower computational cost and are reasonable in complete combustion.

A dynamical storage/deletion algorithm for acceleration of chemical reaction calculation is proposed which deals with stiff chemical source term in reactive flows.The algorithm is applied to two-dimensional gaseous detonation.Two deletion approaches, global deletion and node deletion,are developed and compared with traditional direct integration(DI) algorithm.It shows that the proposed algorithm ensures computational accuracy and relaxes stringent limitation on computer memory.With fixed size of data table,computational speedup factor increases with increament of tolerance error.With same tolerance error,speedup factor of node deletion approach is higher than that of global deletion.The algorithm shows advantages in solving transient reactive flows.

A joint-scalar probability density function (PDF) model is used to simulate bluff-body stabilized turbulent non-premixed Sydney flames HM1. In Situ Adaptive Tabulation (ISAT) algorithm is used to accelerate chemistry calculations. A modified LRR-IP Reynolds stress model is applied to obtain mean flow and turbulent mixing fields. We consider three chemical kinetical schemes for methane chemistry. Simulation results are compared with experimental data. It shows that the model and mechanisms predict flow field, scalar field and local extinction well. C2 chemistry has minor effect on flame HM1.

A power consumption equation of MCM (Multi-Chip Module) interconnect is presented in an S-domain RLC transmission line model. Simulation results are shown to verify the theoretical analysis.

Based on the wave propagation algorithm, a numerical scheme of multi-component reactive system is constructed. The interaction between the incident shock wave and the flame bubble, and between the reflected shock wave and distorted flame of the CH₄/air mixture are investigated numerically in the Cl₄/air elementary reaction model and the splitting method. The evolution and properties of the flame instability induced by incident and reflected shock waves are discussed. The computation indicates that the Helmholtz instability, Richtmyer-Meshkov instability and chemical heat release rate play important roles in the interaction between shock waves and the flame. The calculated results are compared with experimental ones, and an agreement is shown.

A consistent hybrid FV/MC algorithm developed on triangular unstructured meshes is implemented to simulate turbulent diffusion flame stabilized on bluff body. Monte Carlo method is used to solve joint fluctuating velocity-frequency-composition PDF equation, while Reynolds averaged mass, momentum and energy equations are solved by finite volume method. The coupling between the two methods reduces the statistical and bias error of stand-alone particle method, so accuracy and efficiency are improved greatly. Laminar flamelet model is incorporated. The simulation results are compared with the experimental data.