Publications

This paper describes several different types of constraints that can be placed on multilayered feedforward neural networks which are used for automatic target recognition (ATR). We show how unconstrained networks are likely to give poor generalization on the ATR problem. We also show how the ATR problem requires a special type of classifier called a one-class classifier. The network constraints come in two forms: architectural constraints and learning constraints. Some of the constraints are used to improve generalization, while others are incorporated so that the network will be forced to perform one-class classification. 14 refs

Foams, like most highly structured fluids, exhibiting rheological behavior that is both fascinating and complex. We have developed microrheological models for uniaxial extension and simple shearing flow of a `dry`, perfectly ordered, three-dimensional foam composed of thin films with uniform surface tension T and negligible liquid content. We neglect viscous flow in the thin films and examine large elastic-plastic deformations of the foam. The primitive undeformed foam structure is composed of regular space-filling tetrakaidecahedra, which have six square and eight hexagonal surfaces. This structure possesses the film-network topology that is necessary to satisfy equilibrium: three films meet at each edge, which corresponds to a Plateau border, and four edges meet at vertex. However, to minimize surface energy, the films must meet at equal angles of 120{degrees} and the edges must join at equal tetrahedral angles of cos{sup {minus}1}({minus}1/3) {approx} 10.947{degree}. No film in an equilibrium foam structure can be a planar polygon because no planar polygon has all angles equal to the tetrahedral edge. In the equilibrium foam structure known as Kelvin`s minimal tetrakaidecahedron, the `squares` are planar quadrilateral surfaces with curved edges and the `hexagons` are non-planar saddle surfaces with zero mean curvature. As the foam structure evolves with the macroscopic flow, each film maintains zero mean curvature because the pressure is the same in every bubble. In general, the shape of each thin film, defined by z = h(x,y), satisfies R{sub 1}/1 + R{sub 2}/1 = {del}{center_dot} (1 + {vert_bar}{del}h{vert_bar}){sup {1/2}} = O where R{sub 1}{sup {minus}1} and A{sub 2}{sup {minus}1} are the principal curvatures. The appropriate boundary conditions correspond to three films meeting at equal angles. For the homogeneous deformations under consideration, the center of each film moves affinely with the flow. 5 refs

We have measured the atmospheric isoplanatic angle in the near infrared by correlating widely separated time-resolved stellar images.

We have grown Zn{sub 1-x}Mn{sub x}Te alloys by molecular beam epitaxy and characterized them using x-ray diffraction and low temperature magnetoluminescence. Zn{sub 1-x}Mn{sub x}Te is a dilute magnetic semiconductor (DMS) whose bandgap ranges from the green through the blue part of the spectrum and is therefore of interest for blue LEDs.

Renewable energy technologies convert naturally occurring phenomena into useful energy forms. These technologies use resources that generally are not depleted, such as the direct energy (heat and light) from the sun and the indirect results of its impact on the earth (wind, falling water, heating effects, plant growth), gravitational forces (the tides), and the heat of the Earth`s core (geothermal), as the sources from which they produce useful energy. These very large stores of natural energy represent a resource potential that is incredibly massive -- dwarfing that of equivalent fossil energy resources. The magnitude of these resources is, therefore, not a key constraint on energy production. However, they are generally diffuse and not fully accessible, some are intermittent, and all have distinct regional and local variability. It is these aspects of their character that give rise to difficult, but generally solvable, technical, institutional, and economic challenges inherent in development and use of renewable energy resources. This report discusses the technologies and their associated energy source.

Theoretical models have been formulated describing the dynamic behavior of the swelling and contracting of polyelectrolyte gels. This paper presents a method of weighted residuals approach to solving the governing system of equations by finite element analysis. The modulation of the imbibition of solvent by a spherical gel is studied.

There is considerable interest in the use of chemically vapor deposited (CVD) polycrystalline diamond films in advanced materials technology. However, most of the potential applications of CVD diamond films require well-controlled properties which depend on the film structure, and in turn, on the conditions under which the films are synthesized. The structure of the vapor-deposited diamond films is frequently characterized by Raman spectroscopy. Despite extensive research, much work still needs to be completed to understand the various features of the Raman spectra and to understand how the processing variables affect the spectral features. This paper examines the Raman spectra of diamond films prepared by a hot-filament-assisted CVD process as a function of substrate processing and deposition parameters.

Many applications of national importance require the design, analysis, and simulation of complex electromagnetic phenomena. These applications range from the simulation of synthetic aperture radar to the design and analysis of low-observable platforms, antenna design, and automatic target recognition. In general, the modeling of complex electromagnetic phenomena requires significant amounts of computer time and capacity on conventional vector supercomputers but takes far less on massively parallel computers. Sandia National Laboratories is currently developing massively parallel methods and algorithms for the characterization of complex electromagnetic phenomena. The goal of on going research at Sandia is to understand the characteristics, limitations, and trade-offs associated with complex electromagnetic systems including: modeling the seeker response to complex targets in clutter, calculating the radiation and scattering from conformal communication and radar system antennas, and the analysis and design of high speed circuitry. By understanding the theoretical underpinnings of complex electromagnetic systems it is possible to achieve realistic models of system performance. The first objective is the development of computationally practical, high fidelity, systems models targeted for massively parallel computers. Research to achieve this objective is conducted in such areas as mathematical algorithms, problem decomposition, inter-processor communication schemes, and load balancing. The work in mathematical algorithms includes both the development of new methods and the parallel implementation of existing techniques. The second objective is the application of these high fidelity models to facilitate a better understanding of systems level performance for many C{sup 3}I platforms. This presentation describes applications of much current interest and novel solution techniques for these applications utilizing massively parallel processing techniques.

A neighboring external control problem is formulated for a hypersonic glider to execute a maximum-terminal-velocity descent to a stationary target. The resulting two-part, feedback control scheme initially solves a nonlinear algebraic problem to generate a nominal trajectory to the target altitude. Secondly, a neighboring optimal path computation about the nominal provides a lift and side-force perturbations necessary to achieve the target downrange and crossrange. On-line feedback simulations of the proposed scheme and a form of proportional navigation are compared with an off-line parameter optimization method. The neighboring optimal terminal velocity compares very well with the parameter optimization solution and is far superior to proportional navigation. 8 refs.

A camera with 3-ns time resolution -- and a continuous (> 100-ns) record length -- has been developed to image a 10{sup 12}-10{sup 13} W/cm{sup 2} ion beam for inertial-confinement-fusion experiments. A thin gold Rutherford-scattering foil placed in the path of the beam scatters ions into the camera. The foil is in a near-optimized scattering geometry and reduces the beam intensity seven orders of magnitude. The scattered ions are pinhole imaged onto a 2-D array of 39 p-i-n diode detectors. The output of each detector is recorded on a LeCroy 6880 transient-waveform digitizer. The waveforms are analyzed and combined to produce a 39-pixel movie which can be displayed on an image processor to provide, for example, time-resolved horizontal- and vertical-focusing information.

A survivable debris shield is necessary if very high levels of cleanliness are to be achieved on samples exposed to soft x rays from laboratory simulation sources. These samples, often of optical quality, require a low or zero debris environment in order to distinguish low levels of soft x-ray damage from debris-induced effects. We will describe the development of survivable debris shield technology on the Sandia National Laboratories` Saturn x-ray simulation facility, which permits ultraclean exposure over sample areas of 58 cm{sup 2} {at} {approximately}0.35 cal/cm{sup 2}, 5 cm{sup 2}{at} {approximately}0.8 cal/cm{sup 2}, cm{sup 2}, 2.8 cm{sup 2} {at} {approximately}1.2 cal/cm{sup 2}, and 0.8 cm{sup 2} {at} {approximately}1.5 cal/cm{sup 2} with 3-keV argon K-shell photons. These new fluence-area produce test capabilities represent an order-of-magnitude improvement in the state of the art (previously 0.25 cm{sup 2} {at} {approximately} 1 cal/cm{sup 2} or 1.0 cm{sup 2} {at} {approximately}0.35 cal/cm{sup 2}) for soft x-ray simulators.

Methods which report an ability to determine interstitial oxygen (O{sub i}) in heavily doped silicon include the short baseline (SBL) and the curved baseline (CBL) techniques. Both the SBL and CBL methods are similar with the exception of the calibration. We developed a FORTRAN software package to provide separate methods for calculating O{sub i}. The strength of the software is its ability to rapidly compare methods. We report on a comparison of calculation methods including the ASTM standard test F1188-88 which is used in a calibration, and the SBL and the CBL methods, which are used for the heavily doped silicon to provide reliable results. The O{sub i} values calculated for both virgin and thermally processed silicon are reported.

This paper describes the design of an inverse adaptive filter, using the Least-Mean-Square (LMS) algorithm, the correct data taken with an analog filter. The gradient estimate used in the LMS algorithm is based upon the instantaneous error, e{sup 2}(n). Minimizing the mean-squared-error does not provide an optimal solution in this specific case. Therefore, another performance criterion, error power, was developed to calculate the optimal inverse model. Despite using a different performance criterion, the inverse filter converges rapidly and gives a small mean-squared-error. Computer simulations of this filter are also shown in this paper.

Intense light ion beams are being developed to drive inertial confinement fusion (ICF) targets. Recently, intense proton beams have been used to drive two different types of targets in experiments on the Particle Beam Fusion Accelerator. The experiments focused separately on ion deposition physics and on implosion hydrodynamics. In the ion deposition physics experiments, a 3--4 TW/cm{sup 2} proton beam heated a low-density foam contained within a gold cylinder with a specific power deposition exceeding 100 TW/gm for investigating ion deposition, foam heating, and generation of x-rays. The significant results from these experiments included the following: the foam provided an optically thin radiating region, the uniformity of radiation across the foam was good, and the foam tamped the gold case, holding it in its original position for the 15 ns beam pulse width.

This paper identifies some of the sources of resistance to change that face Sandia National Laboratories` management, and examines the mechanisms that are being used to minimize the effects of that resistance. The rapid and radical nature of many of the changes now being demanded of Sandia make them all the more likely to meet stiff resistance from within the organization. Sandia`s management has attempted to anticipate some of the sources of resistance to the changes planned for its operating procedures and has taken action to minimize their effects. Within this paper, particular attention is given to those individuals who will not only be most effected by, by may also be required to implement the changes. The change process at Sandia is currently ongoing. It is not complete. No claim is made that all of the changes that will be necessary at Sandia National Laboratories have been recognized, nor that all sources of resistance have been anticipated.

A production vacuum arc remelt (VAR) furnace was modified to enable direct viewing of the metal vapor arc and molten electrode tip during melting of 432 mm dia. alloy 718 electrodes into 508 mm dia. ingots. Diffuse and constricted arcing conditions were characterized using high speed cinematography, standard video format, and monochromatic imaging. Constricted arcing was observed while melting electrodes contaminated with oxide slag of the type used for refractory linings in vacuum induction furnaces. Monochromatic imaging was used in visualize the ion distribution in the arc plasma; these images clearly showed whether the arc operated in a diffuse or constricted model. Diffuse arc melting conditions were very similar to those previously reported in the literature for smaller laboratory sized melts.

In November of 1989, technology transfer became a mission for Sandia National Laboratories, (SNL), with the passage of the National Competitiveness Technology Transfer Act. In order to address the specialized technology transfer needs of small businesses, SNL created and implemented the Technology-Based Regional Economic Development (TRED) program. The TRED model has two major components -- technology assistance (or teaming), and `` widget transfer.`` In the technology assistance component, SNL`s technology resources (expertise, services, and equipment) are made available to companies developing commercial products. In the ``widget transfer`` component, SNL`s intellectual property (patents, copyrights) is placed with private sector firms through various partnership intermediaries

This manual is intended to aid operators of the pressure controller in the Transducer Calibration Laboratory. We include discussions of automated control, data acquisition, and data analysis techniques that improve the quality and thoroughness of transducer pressure calibration procedures.

This memorandum is a synopsis of the description and operation of the equipment used and the events occurring during the calibration of an accelerometer on the 5000 g centrifuge.

Many MoS{sub 2} lubricant film formulations exhibit tribological performance degradation after extended storage in high humidity ambient environments. As a part of an extensive study of the effects of high humidity storage on a wide variety of MoS{sub 2} lubricant film formulations, the effects of high humidity storage on sputtered MoS{sub 2} films have been examined. The surface chemistries of a series of ten different commercially deposited MoS{sub 2} lubricant film formulations on 440C stainless steel have been examined before and after 1 month of storage at 98% relative humidity. While the films varied greatly in film chemistry prior to high humidity storage, many smaller differences were observed after high humidity storage. Addition of Ni to the films affects the film chemistry far more dramatically than the addition of AuPd or SbO{sub x}, facilitating oxidation of the MoS{sub 2} and enhancing formation of sulfate species. All of the films are adequately characterized by considering only MoS{sub 2} and MoO{sub 3} species, with the exception of the films co-sputtered with Ni and ion implanted with N{sup +}, where some MoO{sub 2} or MoS{sub 2-x}O{sub x} must also be present.

This bibliography contains 64 references concerning the management of business and industry. Books and articles are both cited. Methods for gathering and utilizing information are emphasized. (GHH)

A power MOSFET macro-model for use with the circuit simulator SPICE has been developed suitable for use over the temperature range of {minus}55 to 125{degrees}C. The model is comprised of a single parameter set with the temperature dependence accessed through the SPICE TEMP card. This report describes in detail the development of the model and the extraction algorithms used to obtain model parameters. The extraction algorithms are described in sufficient detail to allow for automated measurements which in turn allows for rapid and cost effective development of an accurate SPICE model for any power MOSFET. 22 refs.

The analytical procedures for establishing the fragment hazard zone for explosive test facilities are presented. Environment, safety and health regulations require that a hazard zone analysis be conducted for every explosive test facility. Analyses are presented for explosively driven missile fragment trajectories resultant from cased explosive configurations. Fragment trajectory parameter data are presented in graphical form for three different fragment materials (aluminum, steel and tantalum), initial velocities between 0.6mm/{mu}s (2000 ft/sec) to 4.3mm/{mu}s (14,000 ft/sec), and for various geometries. This trajectory information is used, as an example, to determine the safe distance or hazard zone for the Area 2 explosive test facility at Sandia National Laboratories.

The equations of motion of beam and channel particles are analyzed in the ion focused regime. Using the paraxial approximation and assuming only transverse electro-magnetostatic interactions between beam and channel particles for the equations of motion (the same equations solved in the BUCKSHOT code) are written in non-dimensional form and scaling relations are derived for propagation parameters, magnetic erosion, and evaporation. These relations are very useful in doing parameter studies with a limited number of computer simulations.

Methods of mechanical and thermal testing specifically directed toward evaluation of impact-limiting materials for radioactive material transportation containers are presented. Associated figures of merit and procedures for rank-ordering the materials are also developed. Based on testing and evaluation procedures developed herein, thermal and mechanical results are presented for high- and low-density aluminum honeycombs, polyurethane foams, and aluminum foams. On a minimum mass basis, the high-density aluminum honeycomb is found to be a superior impact energy absorber up to the point of lock up. On a minimum volume basis, the high-density polyurethane foam is far superior, however. Based on the thermal figures of merit, the high-density polyurethane foam was found to be the most favorable material.