Today, there are several drawbacks that impede the necessary and much needed use of robot learning techniques in real applications. First, the time needed to achieve the synthesis of any behavior is prohibitive. Second, the robot behavior during the learning phase is � by definition � bad, it may even be dangerous. Third, except within the lazy learning approach, a new behavior implies a new learning phase. We propose in this paper to use self-organizing maps to encode the non explicit model of the robot-world interaction sampled by the lazy memory, and then generate a robot behavior by means of situations to be achieved, i.e., points on the self-organizing maps. Any behavior can instantaneously be synthesized by the definition of a goal situation. Its performance will be minimal (not evidently bad) and will improve by the mere repetition of the behavior.

We present results of measurements of spot size and angular divergence of a 30 GeV electron beam through use of optical transition radiation (OTR). The OTR near field pattern and far field distribution are measured as a function of beam spot size and divergence at wavelengths of 441, 532, and 800 nm, for both the single and double foil configurations. Electron beam spot sizes of 50 {micro}m rms have been resolved, demonstrating the utility of OTR for measurement of small beam spot sizes of high energy (30 GeV) electron beams. Two-foil interference was clearly observed and utilized electron beam angular divergences of {approximately} 100 {micro}rad.

A feed-forward multilayer neural net is trained to learn the correspondence between seismic data and well logs. The introduction of a virtual input layer, connected to the nominal input layer through a special nonlinear transfer function, enables ultrafast (single iteration), near-optimal training of the net using numerical algebraic techniques. A unique computer code, named DeepNet, has been developed, that has achieved, in actual field demonstrations, results unattainable to date with industry standard tools.

A method is described for inverse modeling of material deformation in applications of importance to the sheet metal forming industry. The method was developed in order to assess the feasibility of utilizing empirical data in the early stages of the design process as an alternative to conventional prototyping methods. Because properly prepared and employed artificial neural networks (ANN) were known to be capable of codifying and generalizing large bodies of empirical data, they were the natural choice for the application. The product of the work described here is a desktop ANN system that can produce in one pass an accurate die design for a user-specified part shape.

This is the final report of a one-year, Laboratory-Directed Research and Development (LDRD) project at the Los Alamos National Laboratory (LANL). The purpose of this project to study the effects of materials migration by direct measurement of the diffusion and convective migration processes in complex polymeric materials and to develop appropriate predictive models. The use of isotopically tagged probe molecules to measure in-situ the diffusion of water in estane was demonstrated. A special environmental cell with a thin window was fabricated to enable real time measurements to be made under realistic conditions simulating actual operating parameters. Depth profiles were measured quantitatively using ion beam methods available at the Los Alamos Ion Beam Materials Laboratory. The Williams-Landau-Ferry model was adopted as a general expression for diffusion of a volatile material in a polymer. This model contains both thermal activation and free-volume change effects to account for the changes in polymeric structure with temperature and physical properties as embodied in the glass-transition temperature. A theoretical simulation of water migration in polyurethane was performed and compared to the ideal 1-D, constant temperature, constant-boundary concentration test problem, for which an analytical solution is known. The transport code works properly and indicates that time steps …

This is the final report of a one-year, Laboratory-Directed Research and Development (LDRD) project at the Los Alamos National Laboratory (LANL). Methane, the primary component of natural gas, has reserves that are on the order of those of petroleum. Processes that utilize these vast supplies of methane will need to be developed to replace dwindling supplies of petroleum in the future. Processes utilizing natural gas promise to be more environmentally friendly, as natural gas as a feedstock is freer of contaminants and more readily purified than petroleum. Short contact time reactor configurations are likely candidates for this application. The authors objectives are to develop reactor designs and computer models appropriate for short contact time applications. They have succeeded in assembling both an experimental facility for investigating the performance of short contact time reactors, and a computer simulation that includes full mass and heat transport as well as coupled surface and gas phase detailed chemical kinetics.

This is the final report of a two-year, Laboratory Directed Research and Development (LDRD) project at Los Alamos National Laboratory (LANL). The purpose of this project was to develop diamond and diamond-like thin-films as hydrogen isotope permeation barriers. Hydrogen embrittlement limits the life of boost systems which otherwise might be increased to 25 years with a successful non-reactive barrier. Applications in tritium processing such as bottle filling processes, tritium recovery processes, and target filling processes could benefit from an effective barrier. Diamond-like films used for low permeability shells for ICF and HEDP targets were also investigated. Unacceptable high permeabilities for hydrogen were obtained for plasma-CVD diamond-like-carbon films.

This paper describes the development and evaluation of neural network-based systems for industrial resistance spot welding process control and weld quality assessment. The developed systems utilize recurrent neural networks for process control and both recurrent networks and static networks for quality prediction. The first section describes a system capable of both welding process control and real-time weld quality assessment, The second describes the development and evaluation of a static neural network-based weld quality assessment system that relied on experimental design to limit the influence of environmental variability. Relevant data analysis methods are also discussed. The weld classifier resulting from the analysis successfldly balances predictive power and simplicity of interpretation. The results presented for both systems demonstrate clearly that neural networks can be employed to address two significant problems common to the resistance spot welding industry, control of the process itself, and non-destructive determination of resulting weld quality.

Superconductivity can be induced at high temperatures in Pb{sub 2}Sr{sub 2}RCu{sub 3}O{sub 8} (R - rare earth) by partially doping Ca{sup 2+} for R{sup 3+}. In order to understand the interplay between magnetism and superconductivity, the magnetic properties of the parent compounds, Pb{sub 2}Sr{sub 2}RCu{sub 3}O{sub 8}, have been studied. The work presented here includes magnetic susceptibility and specific heat measurements on R=Dy and extends the previous studies on R=Ce, Pr, Tb, Ho and Er. Specific heat experiments suggest that the Dy ions order antiferromagnetically with an ordering temperature of 1.3K. The magnetic susceptibility data are in good agreement with the susceptibility calculated using crystal field parameters that are extrapolated from previous modeling of the R=Er and Ho analogs of this series.