SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Optoelectronic apparatus acts as artificial neural network performing associative recall of binary images. Recall process is iterative one involving optical computation of inner products between binary input vector and one or more reference binary vectors in memory. Inner-product method requires far less memory space than matrix-vector method.
MPERRE computer program accelerates convergence of sequence of vectors by use of minimum-polynomial extrapolation (MPE) and reduced-rank extrapolation (RRE). Effective in accelerating convergences of such sequences of vectors as those obtained from iterative solution of systems of linear and nonlinear equations. In conjunction with various iterative techniques, successfully employed in finite-difference solution of large-scale elliptic boundary-value problems and problems in computational-fluid-dynamics. Only input required is sequence of vectors, convergence of which is accelerated. Program economical and easy to use. Written in FORTRAN 77.
This report summarizes part of the research work accomplished during the second year of a two-year grant. The research, entitled 'Application of Lanczos Vectors to Control Design of Flexible Structures' concerns various ways to use Lanczos vectors and Krylov vectors to obtain reduced-order mathematical models for use in the dynamic response analyses and in control design studies. This report presents a one-sided, unsymmetric block Lanczos algorithm for model reduction of structural dynamics systems with unsymmetric damping matrix, and a control design procedure based on the theory of linear state function observers to design low-order controllers for flexible structures.
Modeling global atmospheric circulations and forecasting the weather would improve greatly if worldwide information on winds aloft were available. Recognition of this led to the inclusion of the LAser Wind Sounder (LAWS) system to measure Doppler shifts from aerosols in the planned for Earth Observation System (EOS). However, gaps will exist in LAWS coverage where heavy clouds are present. The RAdar Wind Sensor (RAWS) is an instrument that could fill these gaps by measuring Doppler shifts from clouds and rain. Previous studies conducted at the University of Kansas show RAWS as a feasible instrument. This thesis pertains to the signal-to-noise ratio (SNR) sensitivity, transmit waveform, and limitations to the antenna scan pattern of the RAWS system. A dop-size distribution model is selected and applied to the radar range equation for the sensitivity analysis. Six frequencies are used in computing the SNR for several cloud types to determine the optimal transmit frequency. the results show the use of two frequencies, one higher (94 GHz) to obtain sensitivity for thinner cloud, and a lower frequency (24 GHz) to obtain sensitivity for thinner cloud, and a lower frequency (24 GHz) for better penetration in rain, provide ample SNR. The waveform design supports covariance estimation processing. This estimator eliminates the Doppler ambiguities compounded by the selection of such high transmit frequencies, while providing an estimate of the mean frequency. the unambiguous range and velocity computation shows them to be within acceptable limits. The design goal for the RAWS system is to limit the wind-speed error to less than 1 ms(exp -1). Due to linear dependence between vectors for a three-vector scan pattern, a reasonable wind-speed error is unattainable. Only the two-vector scan pattern falls within the wind-error limits for azimuth angles between 16 deg to 70 deg. However, this scan only allows two components of the wind to be determined. As a result, a technique is then shown, based on the Z-R-V relationships, that permit the vertical component (i.e., rain) to be computed. Thus the horizontal wind components may be obtained form the covariance estimator and the vertical component from the reflectivity factor. Finally, a new candidate system is introduced which summarizes the parameters taken from previous RAWS studies, or those modified in this thesis.
This thesis describes the design and functionality of a Distributed Petri Net Controller (DPNC). The controller runs under X Windows to provide a graphical interface. The DPNC allows users to distribute a Petri Net across several host computers linked together via a TCP/IP interface. A sub-net executes on each host, interacting with the other sub-nets by passing a token vector from host to host. One host has a command window which monitors and controls the distributed controller. The input to the DPNC is a net definition file generated by Great SPN. Thus, a net may be designed, analyzed and verified using this package before implementation. The net is distributed to the hosts by tagging transitions that are host-critical with the appropriate host number. The controller will then distribute the remaining places and transitions to the hosts by generating the local nets, the local marking vectors and the global marking vector. Each transition can have one or more preconditions which must be fulfilled before the transition can fire, as well as one or more post-processes to be executed after the transition fires. These implement the actual input/output to the environment (machines, signals, etc.). The DPNC may also be used to simulate a Great SPN net since stochastic and deterministic firing rates are implemented in the controller for timed transitions.
The relationship between the sequential hard c-means (SHCM), learning vector quantization (LVQ), and fuzzy c-means (FCM) clustering algorithms is discussed. LVQ and SHCM suffer from several major problems. For example, they depend heavily on initialization. If the initial values of the cluster centers are outside the convex hull of the input data, such algorithms, even if they terminate, may not produce meaningful results in terms of prototypes for cluster representation. This is due in part to the fact that they update only the winning prototype for every input vector. The impact and interaction of these two families with Kohonen's self-organizing feature mapping (SOFM), which is not a clustering method, but which often leads ideas to clustering algorithms is discussed. Then two generalizations of LVQ that are explicitly designed as clustering algorithms are presented; these algorithms are referred to as generalized LVQ = GLVQ; and fuzzy LVQ = FLVQ. Learning rules are derived to optimize an objective function whose goal is to produce 'good clusters'. GLVQ/FLVQ (may) update every node in the clustering net for each input vector. Neither GLVQ nor FLVQ depends upon a choice for the update neighborhood or learning rate distribution - these are taken care of automatically. Segmentation of a gray tone image is used as a typical application of these algorithms to illustrate the performance of GLVQ/FLVQ.
A method of dynamic substructuring is presented which utilizes a set of static Ritz vectors as a replacement for normal eigenvectors in component mode synthesis. This set of Ritz vectors is generated in a recurrence relationship, which has the form of a block-Krylov subspace. The initial seed to the recurrence algorithm is based on the boundary flexibility vectors of the component. This algorithm is not load-dependent, is applicable to both fixed and free-interface boundary components, and results in a general component model appropriate for any type of dynamic analysis. This methodology was implemented in the MSC/NASTRAN normal modes solution sequence using DMAP. The accuracy is found to be comparable to that of component synthesis based upon normal modes. The block-Krylov recurrence algorithm is a series of static solutions and so requires significantly less computation than solving the normal eigenspace problem.
Original Backprop (Version 1.2) is an MS-DOS package of four stand-alone C-language programs that enable users to develop neural network solutions to a variety of practical problems. Original Backprop generates three-layer, feed-forward (series-coupled) networks which map fixed-length input vectors into fixed length output vectors through an intermediate (hidden) layer of binary threshold units. Version 1.2 can handle up to 200 input vectors at a time, each having up to 128 real-valued components. The first subprogram, TSET, appends a number (up to 16) of classification bits to each input, thus creating a training set of input output pairs. The second subprogram, BACKPROP, creates a trilayer network to do the prescribed mapping and modifies the weights of its connections incrementally until the training set is leaned. The learning algorithm is the 'back-propagating error correction procedures first described by F. Rosenblatt in 1961. The third subprogram, VIEWNET, lets the trained network be examined, tested, and 'pruned' (by the deletion of unnecessary hidden units). The fourth subprogram, DONET, makes a TSR routine by which the finished product of the neural net design-and-training exercise can be consulted under other MS-DOS applications.
We present the results of a study of the peculiar velocities of rich clusters of galaxies. The peculiar motion of rich clusters in various cosmological scenarios is of interest for a number of reasons. Observationally, one can measure the peculiar motion of clusters to greater distances than galaxies because cluster peculiar motions can be determined to greater accuracy. One can also test the slope of distance indicator relations using clusters to see if galaxy properties vary with environment. We have used N-body simulations to measure the amplitude and rms cluster peculiar velocity as a function of bias parameter in the hot and cold dark matter scenarios. In addition to measuring the mean and rms peculiar velocity of clusters in the two models, we determined whether the peculiar velocity vector of a given cluster is well aligned with the gravity vector due to all the particles in the simulation and the gravity vector due to the particles present only in the clusters. We have investigated the peculiar velocities of rich clusters of galaxies in the cold dark matter and hot dark matter galaxy formation scenarios. We have derived peculiar velocities and associated errors for the scenarios using four values of the bias parameter ranging from b = 1 to b = 2.5. The growth of the mean peculiar velocity with scale factor has been determined and compared to that predicted by linear theory. In addition, we have compared the orientation of force and velocity in these simulations to see if a program such as that proposed by Bertschinger and Dekel (1989) for elliptical galaxy peculiar motions can be applied to clusters. The method they describe enables one to recover the density field from large scale redshift distance samples. The method makes it possible to do this when only radial velocities are known by assuming that the velocity field is curl free. Our analysis suggests that this program if applied to clusters is only realizable for models with a low value of the bias parameter, i.e., models in which the peculiar velocities of clusters are large enough that the errors do not render the analysis impracticable.
The principal objective is to provide a training procedure for a feed forward, back propagation neural network which greatly accelerates the training process. A set of orthogonal singular vectors are determined from the input matrix such that the standard deviations of the projections of the input vectors along these singular vectors, as a set, are substantially maximized, thus providing an optimal means of presenting the input data. Novelty exists in the method of extracting from the set of input data, a set of features which can serve to represent the input data in a simplified manner, thus greatly reducing the time/expense to training the system.
The cultivation of irrigated rice provides ideal larval habitat for a number of anopheline vectors of malaria throughout the world. Anopheles freeborni, a potential vector of human malaria, is associated with the nearly 240,000 hectares of irrigated rice grown annually in Northern and Central California; therefore, this species can serve as a model for the study of rice field anopheline population dynamics. Analysis of field data revealed that rice fields with early season canopy development, that are located near bloodmeal sources (i.e., pastures with livestock) were more likely to produce anopheline larvae than fields with less developed canopies located further from pastures. Remote sensing reflectance measurements of early-season canopy development and geographic information system (GIS) measurements of distanes between rice fields and pastures with livestock were combined to distinguish between high and low mosquito-producing rice fields. Using spectral and distance measures in either a discriminant or Bayesian analysis, the identification of high mosquito-producing fields was made with 85 percent accuracy nearly two months before anopheline larval populations peaked. Since omission errors were also minimized by these approaches, they could provide a new basis for directing abatement techniques for the control of malaria vectors.
The hardware and algorithms used to vector quantize (VQ) predicted pixel intensity differences for real-time video compression are described. The hardware is designed for rapid vector quantization performance, which entails the development of application-specific associative memory circuits. A modified DPCM algorithm is originally examined to determine how neural circuitry could enhance its operation. It was determined that quantization and encoding could be improved by consolidating these two functions into one, and by increasing the amount of information (i.e. number of pixels) quantized at a time. The result is a predictive scheme that vector quantizes differential values. Some of the disadvantages of VQ algorithms are solved using associative memories. The video compression algorithm and the associative memory design are described.
Magnetic field and fluid vorticity share many features. First, as divergence-free vector fields they are conveniently visualized in terms of their field lines, curves that are everywhere tangent to the field. The lines indicate direction and their density indicates field strength. The question arises of the extent to which the evolution of the fields can be treated in terms of the evolution of their field lines. Newcomb (1958) derived the general conditions on the evolution of vector fields that permit the identification of field lines from one instant to the next. The equations of evolution of the vorticity field and the magnetic field fall within Newcomb's analysis. The dynamics of the flows differ between these two systems, so that geometrically similar phenomena happen in different ways in the two systems. In this paper the geometrical similarities are emphasized. Reconnection will be defined here as evolution in which it is not possible to preserve the global identification of some field lines. There is a close relation between reconnection and the topology of the vector field lines. Nontrivial topology occurs where the field has null points or there are field lines that are closed loops.
We present a study of the visualization of 3-D second order tensor fields and matrix data. The general problem of visualizing unsymmetric real or complex Hermitian second order tensor fields can be reduced to the simultaneous visualization of a real and symmetric second order tensor field and a real vector field. As opposed to the discrete iconic techniques commonly used in multivariate data visualization, the emphasis is on exploiting the mathematical properties of tensor fields in order to facilitate their visualization and to produce a continuous representation of the data. We focus on interactively sensing and exploring real and symmetric second order tensor data by generalizing the vector notion of streamline to the tensor concept of hyperstreamline. We stress the importance of a structural analysis of the data field analogous to the techniques of vector field topology extraction in order to obtain a unique and objective representation of second order tensor fields.
When launching a spacecraft from Earth parking orbit to deep space, it is highly desirable to have the hyperbolic excess velocity vector (v-infinity) contained in the parking orbit plane. Ground launches can force the parking orbit plane to contain the v-infinity vector by using launch azimuth and lift-off time as independent variables. When launching from the Space Station, a new set of variables comes into play. The Space Station orbit is of fixed inclination but precessing due to the Earth's oblateness. Its plane will seldom (and may never) contain the desired v-infinity vector. Consequently, the departure strategy will usually require multiple burns and include a plane change. Also, the concept of 'launch window' will be somewhat different from Earth surface launches. An analysis of the deployment of interplanetary spacecraft from the Space Station is described, with emphasis on the effect of the trajectory characteristics on station operations. Several planetary mission types are analyzed, including manned Mars missions and unmanned high declination departures. The constraint of Space Station orbit nodal position is quantified and the operational implications for Space Station reboost strategy are examined.
This paper describes a technique for obstacle detection, based on the expansion of the image-plane projection of a textured object, as its distance from the sensor decreases. Information is conveyed by vectors whose components represent first-order temporal and spatial derivatives of the image intensity, which are related to the time to collision through the local divergence. Such vectors may be characterized as patterns corresponding to 'safe' or 'dangerous' situations. We show that essential information is conveyed by single-bit vector components, representing the signs of the relevant derivatives. We use two recently developed, high capacity classifiers, employing neural learning techniques, to recognize the imminence of collision from such patterns.
This paper describes a highly interactive method for computer visualization of the basic physical process of dispersion and mixing of fluid elements in convection-diffusion systems. It is based on transforming the vector field from a traditionally Eulerian reference frame into a Lagrangian reference frame. Fluid elements are traced through the vector field for the mean path as well as the statistical dispersion of the fluid elements about the mean position by using added scalar information about the root mean square value of the vector field and its Lagrangian time scale. In this way, clouds of fluid elements are traced and are not just mean paths. We have used this method to visualize the simulation of an industrial incinerator to help identify mechanisms for poor mixing.
Probe pattern compensation is essential in near-field scanning geometry, where there is a great need to accurately know far-field patterns at wide angular range. This paper focuses on a novel formulation and computer simulation to determine the precise need for and effect of probe compensation in cylindrical near-field scanning. The methodology is applied to a linear test array antenna and the NASA scatterometer radar antenna. The formulation is based on representing the probe by its equivalent tangential magnetic currents. The interaction between the probe equivalent aperture currents and the test antenna fields is obtained with the application of a reciprocity theorem. This allows us to obtain the probe vector output pickup integral which is proportional to the amplitude and phase of the electric field induced in the probe aperture with respect to its position to the test antenna. The integral is evaluated for each probe position on the required sampling point on a cylindrical near-field surface enclosing the antenna. The use of a hypothetical circular-aperture probe with a different radius permits us to derive closed-form expressions for its far-field radiation patterns. These results, together with the probe vector output pickup, allow us to perform computer simulated synthetic measurements. The far-field patterns of the test antenna are formulated based on cylindrical wave expansions of both the probe and test antenna fields. In the limit as the probe radius becomes very small, the probe vector output is the direct response of the near-field at a point, and no probe compensation is needed. Useful results are generated to compare the far-field pattern of the test antenna constructed from the knowledge of the simulated near-field with and without probe pattern compensation and the exact results. These results are important since they clearly illustrate the angular range over which probe compensation is needed. It has been found that a probe with an aperture radius of 0.25(lambda), 0.5(lambda), and 1(lambda) needs a little probe compensation, if any, near the test antenna main beam. In addition, a probe with low directivity may provide a better signal-to-noise ratio than a highly directive one. This is evident in test antenna patterns without probe compensation at wide angles.