Augmented Lagrangian Algorithms Based on the Spectral Projected Gradient Method for Solving Nonlinear Programming Problems

The spectral projected gradient method SPG is an algorithm for large-scale bound-constrained optimization introduced recently by Birgin, Martínez, and Raydan. It is based on the Raydan unconstrained generalization of the Barzilai-Borwein method for quadratics. The SPG algorithm turned out to be surprisingly effective for solving many large-scale minimization problems with box constraints. Therefore, it is natural to test its perfomance for solving the sub-problems that appear in nonlinear programming methods based on augmented Lagrangians. In this work, augmented Lagrangian methods which use SPG as the underlying convex-constraint solver are introduced (ALSPG) and the methods are tested in two sets of problems. First, a meaningful subset of large-scale nonlinearly constrained problems of the CUTE collection is solved and compared with the perfomance of LANCELOT. Second, a family of location problems in the minimax formulation is solved against the package FFSQP.     

A Scalable Process for Production of Single-walled Carbon Nanotubes (SWNTs) by Catalytic Disproportionation of CO on a Solid Catalyst

Existing single-walled carbon nanotube synthesis methods are not easily scalable, operate under severe conditions, and involve high capital and operating costs. The current cost of SWNT is exceedingly high. A catalytic method of synthesis has been developed that has shown potential advantages over the existing methods. This method is based on a catalyst formulation that inhibits the formation of undesired forms of carbon; it can be scaled-up and may result in lower production costs.     

大规模声学边界元法的GPU并行计算

提出一种大规模声学边界元法的高效率、高精度GPU并行计算方法.基于Burton-Miller边界积分方程,推导适于GPU的并行计算格式并实现了传统边界元法的GPU加速算法.为提高原型算法的效率,研究GPU数据缓存优化方法.由于GPU的双精度浮点运算能力较低,为了降低数值误差,研究基于单精度浮点运算实现的doublesingle精度算法.数值算例表明,改进的算法实现了最高89.8%的GPU使用效率,且数值精度与直接使用双精度数相当,而计算时间仅为其1/28,显存消耗也仅为其一半.该方法可在普通PC机(8GB内存,NVIDIA Ge Force 660 Ti显卡)上快速完成自由度超过300万的大规模声学边界元分析,计算速度和内存消耗均优于快速边界元法.     

滑移边界对二维Rayleigh-Bénard热对流流态和传热特性的影响

采用并行直接数值模拟（PDM-DNS）计算无滑移和滑移边界二维Rayleigh-Bénard热对流.与无滑移边界形成的随机羽流运动的湍流热对流不同,滑移边界热对流最终形成湍流特征消失,且温度仅分布于四壁的一个大尺度环流的流动形态.平均场近底板的温度分布特性,无滑移边界逐渐变化而滑移边界出现过冲现象.宽高比Γ=1时,Nusselt数（Nu）随Rayleigh数（Ra）的变化具有相同标度指数,Nu~Ra^0.3.滑移边界热对流具有传热增强作用.滑移边界热对流Nu随Γ变化明显,并分为两个阶段,在Γ=0.5时出现Nu_max≈250,是无滑移边界热流Nu的5倍.     

基于输出反馈的一类大型互联非线性系统的鲁棒全局指数稳定

研究基于输出反馈的一类新的大型互联非线性不确定系统的鲁棒全局指数稳定问题,通过构造每个子系统收敛的状态观测器,并对观测器的状态作线性变换,得到鲁棒分散输出反馈控制器.当该反馈控制律作用于该系统时,闭环系统是全局指数稳定的.     

大尺度三维几何尺寸立体视觉测量系统实现

基于立体视觉的大尺度三维几何尺寸测量技术是对外形不规则的大型物体进行几何量测量的有效方法。与传统的视觉测量技术不同,大尺度视觉测量的精度随测量距离的增加而迅速下降,同时大尺度空间下照明条件难以控制,也使得测量变得不稳定。因此如何在大测量尺度的前提下提高测量精度成为这一领域的研究难点。从立体视觉测量系统的原理出发,分析了影响系统精度的各种因素,将相位一致性变换与极线约束条件引入结构光光条中心提取之中,显著提高了图像分析的精度。研制了测量系统的样机并给出了实验结果,现场测试表明该系统可以有效地保证大尺度条件下的测量精度。     

计算机控制光学表面成形中大规模驻留时间求解

采用基于稀疏矩阵的大规模非负最小二乘法,对大口径、微浮雕结构光学元件加工中的驻留时间进行了分析与求解,并对该算法开展了正则化研究。仿真结果表明:与传统非负最小二乘法相比,基于稀疏矩阵的大规模非负最小二乘法精度高、效率快。采用该算法仿真加工平均振幅为1.177 6倍波长的大口径、微浮雕结构光学元件,误差面形均方根收敛至0.067倍波长。     

Impact of Partial Separability on Large-Scale Optimization

ELSO is an environment for the solution oflarge-scale optimization problems. With ELSO the user is required to provide only code for the evaluation of a partially separable function. ELSO exploits the partialseparability structure of the function to computethe gradient efficiently using automatic differentiation.We demonstrate ELSO's efficiency by comparing thevarious options available in ELSO.Our conclusion is that the hybrid option in ELSOprovides performance comparable to the hand-coded option, while having the significantadvantage of not requiring a hand-coded gradient orthe sparsity pattern of the partially separable function.In our test problems, which have carefully coded gradients,the computing time for the hybrid AD option is within a factor of two of thehand-coded option.     

Two-phase flow in heterogeneous porous media III: Laboratory experiments for flow parallel to a stratified system

Two-phase flow in stratified porous media is a problem of central importance in the study of oil recovery processes. In general, these flows are parallel to the stratifications, and it is this type of flow that we have investigated experimentally and theoretically in this study. The experiments were performed with a two-layer model of a stratified porous medium. The individual strata were composed of Aerolith-10, an artificial: sintered porous medium, and Berea sandstone, a natural porous medium reputed to be relatively homogeneous. Waterflooding experiments were performed in which the saturation field was measured by gamma-ray absorption. Data were obtained at 150 points distributed evenly over a flow domain of 0.1 × 0.6 m. The slabs of Aerolith-10 and Berea sandstone were of equal thickness, i.e. 5 centimeters thick. An intensive experimental study was carried out in order to accurately characterize the individual strata; however, this effort was hampered by both local heterogeneities and large-scale heterogeneities.The theoretical analysis of the waterflooding experiments was based on the method of large-scale averaging and the large-scale closure problem. The latter provides a precise method of discussing the crossflow phenomena, and it illustrates exactly how the crossflow influences the theoretical prediction of the large-scale permeability tensor. The theoretical analysis was restricted to the quasi-static theory of Quintard and Whitaker (1988), however, the dynamic effects described in Part I (Quintard and Whitaker 1990a) are discussed in terms of their influence on the crossflow.Roman Letters A interfacial area between the -region and the -region contained within V, m² - a vector that maps onto , m - b vector that maps onto , m - b vector that maps onto , m - B second order tensor that maps onto , m² - C second order tensor that maps onto , m² - E energy of the gamma emitter, keV - f fractional flow of the -phase - g gravitational vector, m/s² - h characteristic length of the large-scale averaging volume, m - H height of the stratified porous medium , m - i unit base vector in the x-direction - K local volume-averaged single-phase permeability, m² - K - {K}, large-scale spatial deviation permeability - { K} large-scale volume-averaged single-phase permeability, m² - K ^* large-scale single-phase permeability, m² - K ^** equivalent large-scale single-phase permeability, m² - K local volume-averaged -phase permeability in the -region, m² - K local volume-averaged -phase permeability in the -region, m² - K - {K } , large-scale spatial deviation for the -phase permeability, m² - K ^* large-scale permeability for the -phase, m² - l thickness of the porous medium, m - l characteristic length for the -region, m - l characteristic length for the -region, m - L length of the experimental porous medium, m - characteristic length for large-scale averaged quantities, m - n outward unit normal vector for the -region - n outward unit normal vector for the -region - n unit normal vector pointing from the -region toward the -region (n = - n ) - N number of photons - p pressure in the -phase, N/m² - p ₀ reference pressure in the -phase, N/m² - local volume-averaged intrinsic phase average pressure in the -phase, N/m² - large-scale volume-averaged pressure of the -phase, N/m² - large-scale intrinsic phase average pressure in the capillary region of the -phase, N/m² - - , large-scale spatial deviation for the -phase pressure, N/m² - p_c , capillary pressure, N/m² - p _c capillary pressure in the -region, N/m² - p capillary pressure in the -region, N/m² - {p _c } ^c large-scale capillary pressure, N/m² - q -phase velocity at the entrance of the porous medium, m/s - q -phase velocity at the entrance of the porous medium, m/s - S_wi irreducible water saturation - S /, local volume-averaged saturation for the -phase - S _i initial saturation for the -phase - S _r residual saturation for the -phase - S ^* { }^*/}^*, large-scale average saturation for the -phase - S saturation for the -phase in the -region - S saturation for the -phase in the -region - t time, s - v -phase velocity vector, m/s - v local volume-averaged phase average velocity for the -phase, m/s - {v } large-scale averaged velocity for the -phase, m/s - v local volume-averaged phase average velocity for the -phase in the -region, m/s - v local volume-averaged phase average velocity for the -phase in the -region, m/s - v -{v } , large-scale spatial deviation for the -phase velocity, m/s - v -{v } , large-scale spatial deviation for the -phase velocity in the -region, m/s - v -{v } , large-scale spatial deviation for the -phase velocity in the -region, m/s - V large-scale averaging volume, m³ - y position vector relative to the centroid of the large-scale averaging volume, m - {y}^c large-scale average of y over the capillary region, m Greek Letters local porosity - local porosity in the -region - local porosity in the -region - local volume fraction for the -phase - local volume fraction for the -phase in the -region - local volume fraction for the -phase in the -region - {}^* { }^*+{ }^*, large-scale spatial average volume fraction - { }^* large-scale spatial average volume fraction for the -phase - mass density of the -phase, kg/m³ - mass density of the -phase, kg/m³ - viscosity of the -phase, N s/m² - viscosity of the -phase, Ns/m² - V /V , volume fraction of the -region ( + =1) - V /V , volume fraction of the -region ( + =1) - attenuation coefficient to gamma-rays, m^-1 - -      

Solving nonlinear integer programs with large-scale optimization software

This paper describes recent experience in tackling large nonlinear integer programming problems using the MINOS large-scale optimization software. A technique is presented for extending the constrained search approach used in MINOS to exploring integer-feasible solutions once a continuous optimal solution is obtained. Computational experience with this approach is described for two classes of problems: quadratic assignment problems and pipeline network design problems.