perform in a predictably safe manner, even in severe abnormal environments. This philosophy has been scrutinized and tested during the intervening years, and has proved successful in practice. A requirement guiding the development of the philosophy was that the resulting design must be simple enough to be amenable to analysis. The inherent simplicity is a safety attribute, because complex analyses, such as those represented by fault trees containing hundreds of branches, are extremely susceptible to error. There are many examples where such errors led analysts to believe systems were safe when they were not, with disastrous consequences. The purpose of this workshop problem is to determine whether the principles developed to ensure hardware safety are applicable in any way to safety-critical software systems. It is possible that hardware associations with software will need to be considered, but whether or not this is true is left as an aspect of the investigation. In order to put the ground rules in perspective, it will be necessary to establish some framework. (Contract DE-AC06-76RL-01830) (DE90-016291; PNL-SA-18090; CONF-900793-4) Avail: NTIS HC/MF A03 Subsequent to developing the conceptual description of a intelligent software system a representation scheme is selected. The knowledge representation scheme constrains the way in which knowledge is represented and executed in a software system. The representation scheme consists of methods for writing down information, organizing information, controlling software execution, and performing inference. This paper discusses a knowledge representation scheme that we are developing for use in building intelligent systems that assist in physical system fault diagnosis. This scheme consists of function and object hierarchies, generalized inference methodologies, and a control scheme that allows for concurrent reasoning during fault diagnosis. DOE N91-10632# Allied-Signal Aerospace Co., Kansas City, MO. J. R. Bradford Aug. 1990 46 p (DE90-016430; KCP-613-4172) Avail: NTIS HC/MF A03 Software was conceived that provided two-way CAD data exchange between AuotCAD's DXF format and the IGES format. Obstacles were found with a newly emerging DXF format, and enhancements were added to accommodate multiple DXF forms. Successful code was written using the C programming language, thus providing portability to other hardware. Future enhancements to the code were identified. DOE N91-10633*# Cornell Univ., Ithaca, NY. Dept. of Computer Mesaac Makpangou and Ken Birman Oct. 1990 27 p (Contract NAG2-593; MDA972-88-C-0024) (NASA-CR-187333; NAS 1.26:187333; TR-90-1165) Avail: NTIS HC/MF A03 CSCL 09B Progress in methodologies for developing robust local area network software has not been matched by similar results for wide area settings. The design of application software spanning multiple local area environments is examined. For important classes of applications, simple design techniques are presented that yield fault tolerant wide area programs. An implementation of these techniques as a set of tools for use within the ISIS system is described. Author N91-10634# Bertin et Cie., Plaisir (France). PROTHE: A SOFTWARE AS A HELP FOR THE STUDIES IN The PROTHE software presented aims to give a data processing aid to the industry which provides a better management of their technical knowledge, and offers simultaneous good conditions for filing, accessing, consulting, and increasing the data base. It takes part in the dissemination of the results of scientific research within companies in the thermal engineering field. The data base is situated in the central part of the software. It is divided into libraries, each of them being dedicated to a specific technical theme. The industry will also be able to acquire libraries prepared by scientific bodies, such as laboratories and technical centers. Such a library is being developed under the theme of heat exchange thermal engineering. ESA N91-10635# Center for Mathematics and Computer Science, Amsterdam (Netherlands). Dept. of Operations Research, Statistics, and System Theory. OPTIMIZATION OF POLLING SYSTEMS O. J. Boxma, H. Levy, and J. A. Westrate (Tiburg Univ., Netherlands) Dec. 1989 20 p (CWI-BS-R8932; ETN-90-97665) Copyright Avail: NTIS HC/MF A03 The issue of deriving efficient operational rules for polling systems with switch-over periods is addressed. The static optimization problem of determining the server visit order (polling table) that minimizes the mean total workload is studied. This problem is strongly related with, and in many applications coincides with, that of minimizing the overall mean customer delay in the system. A heuristic approach to the polling table problem is presented, using the exact solution of a related problem, specifically that of determining those server routing probabilities in a random polling system, that lead to the minimal mean total workload. Numerical experiments show that this heuristic approach yields excellent results. ESA N91-10636# Technische Univ., Delft (Netherlands). Faculty of Technical Mathematics and Informatics. A POTENTIAL REDUCTION METHOD FOR A CLASS OF D. denHertog, C. Roos, and T. Terlaky (Eoetvoes Lorand Univ., A potential reduction method for smooth convex programming is proposed. It is assumed that the objective and constraint functions fulfill the so called relative Lipschitz condition, with Lipschitz constant M greater than 0. A great advantage of the method, above the existing path following methods, is that the steps can be made long by performing linesearches. In the method linesearches are done along the Newton direction with respect to a strictly convex potential function when far away from the central path. When sufficiently close to this path, a lower bound is updated for the optimal value. The number of iterations required by the algorithm to converge to an epsilon optimal solution 0((1+(M squared)) square root of(n)In epsilon) or 0((1+(M_squared)) n In epsilon), dependent on the updating scheme for the lower bound. ESA N91-10637# Technische Univ., Delft (Netherlands). Faculty of ON THE ROTATION INTERPOLATION AND ANIMATION J. E. Mebius 1990 18 p (Rept-90-08; ISSN-0922-5641; ETN-90-97704) Copyright Avail: NTIS HC/MF A03 A new technique for performing three-dimensional rotation interpolation and animation is explained. Its advantages and N91-10638# Technische Univ., Delft (Netherlands). Faculty of Technical Mathematics and Informatics. AUTOMATED SUPPORT OF THE MODELLING PROCESS: A VIEW BASED ON EXPERIMENTS WITH EXPERT INFORMATION ENGINEERS G. M. Winters 1990 22 p Sponsored in part by Dutch User Group of Structured Development Methods; BSO, Rotterdam, Netherlands; Hightech Automation; Nixdorf Computer; and Unisys Corp. (Rept-90-09; ISSN-0922-5641; ETN-90-97705) Copyright Avail: NTIS HC/MF A03 A trend can be discerned clearly indicating that today's developments in methodologies concentrate on an ongoing structuring and integration of modeling techniques. Automated support is the prime mover of this trend. An argument that a lack of explicit knowledge about the process of model construction has resulted in problems with computer aided software engineering tools concerning adequate support in the form of verification and navigation is given. Important aspects of the modeling process deduced from the analysis of experiments with expert information engineers are presented. Some of the important notions are strategy, refinement, division and integration as far as the process of model construction is concerned, and the notions of natural level of consistency and finer categorization of concepts as far as the model constructs themselves are concerned. An architecture of modeling support systems and a corresponding prototype, that are able to reflect the mentioned aspects of modeling processes, are described. ESA SILAGE TO ELLA TRANSLATION M. G. Hill and E. V. Whiting Apr. 1990 25 p SILAGE is an applicative behavioral specification language especially suited for digital signal processing (DSP) applications. A study of the translation of SILAGE circuits to ELLA is carried out. The ELLA system is an integrated hardware design tool set, which comprises the ELLA language compiler, the ELLA application support environment, the ELLA simulator, and the ELLANET procedural interface. A series of ELLA macrofunctions which could be combined into a library for use in the translation of SILAGE circuits is outlined. ESA N91-10640# Leiden Univ. (Netherlands). Div. of Applied DISCOUNTED DYNAMIC PROGRAMMING. PART 3: A. Hordijk and L. C. M. Kallenberg Dec. 1989 26 p The Markov decision theory, with emphasis on theory and methods, was studied. In the method of successive approximation the value vector is approximated by a sequence which converges to it. A nearly optimal policy is determined. The algorithm is presented. Exclusion of suboptimal actions and advanced algorithms are considered. ESA N91-10641# Leiden Univ. (Netherlands). Div. of Applied Mathematics. DISCOUNTED DYNAMIC PROGRAMMING. PART 4: ACCELERATED SUCCESSIVE APPROXIMATION A. Hordijk and L. C. M. Kallenberg Jan. 1990 35 p The Markov decision theory, with emphasis on theory and methods, was studied. Some variants of the successive approximation iterative procedure to solve the functional equation Ux = x are discussed. The pre-Gauss-Seidel, Gauss-Seidel, and the successive over-relaxation techniques are presented and these variants are generalized by using stopping times. All variants are based on contraction mappings and may be considered as an acceleration of the basic algorithm. ESA N91-10642# Leiden Univ. (Netherlands). Dept. of Mathematics and Computer Science. DISCOUNTED DYNAMIC PROGRAMMING. PART 5: MODIFIED A. Hordijk and L. C. M. Kallenberg Jan. 1990 25 p The Markov decision theory, with emphasis on theory and methods was studied. An algorithm for policy improvement was discussed. The algorithm is presented. The convergence rate is examined and bounds on the value vector are given in order to exclude suboptimal actions. ESA N91-10643# Universiteit Twente, Enschede (Netherlands). Tele-Informatics Group. Jan Tretmans Apr. 1990 17 P Sponsored by EEC and Netherlands Ministry of Education (Memo-INF-90-21; ISSN-0923-1714; ETN-90-97731) Avail: NTIS HC/MF A03 Test cases are derived from LOTOS behavior expressions. Starting point is the conformance relation 'conf' and the related theory to derive canonical testers for testing conformance of implementations to specifications. This theory is applicable to formalisms that have labelled transition systems with internal steps as a model. The algorithm for the derivation of canonical testers was already applied to basic LOTOS. Here it is adapted in order to derive test cases from full LOTOS expressions, i.e., LOTOS expressions with data-value communication. In order to apply the derived algorithm, value communication is interpreted as infinite choice over all possible values of the communication. Examples of test cases for simple expressions are given, together with problems and (im)possibilities in this approach. ESA N91-10644# Oxford Univ. (England). Dept. of Engineering Science. CASUAL EIGENVECTOR FUNCTION APPROXIMATIONS AND J. A. Rossiter and B. Kouvaritakis 3 Apr. 1990 19 p Eigenvector scaling implies the existence of degrees of freedom in the power series approximations of the eigenvector and dual eigenvector functions of discrete time transfer function matrix. An earlier study gave consideration to the anticlockwise winding associated with anti-causality and proposed an algorithm which scaled eigenvectors with a view to reducing their anti-causal component. A different algorithm is proposed which, although still formulated in the frequency domain, places a direct penalty on the time domain anti-causal component of the eigenvector sequences and avoids some of the aliasing difficulties that could arise in connection with the earlier algorithm. The new algorithm also provides the systematic means of reducing the length of the causal component of the associated vector sequences and results in better eigenvector approximations, as illustrated by means of two numerical examples. ESA SOME DEFECTS IN FINITE-DIFFERENCE EDGE FINDERS Margaret M. Fleck 1990 27 p Sponsored by British Petroleum Co. Ltd. (OUEL-1826/90; ETN-90-97923) Avail: NTIS HC/MF A03 A study of how performance of finite difference edge finders varies with which order differences they use and how they combine differences taken in different directions is presented. Five algorithms are considered: Canny's, Boie and Cox's, and three variations on Marr and Hildreth's. These algorithms were re-implemented using a uniform method of noise suppression and a uniform output representation. Small problems in the original designs or published descriptions are identified and repaired. All the algorithms produce artifacts on simple test patterns which contains features such as sharp corners, junctions, and intensity slopes. The artifacts include gaps in boundaries, spurious boundaries, and deformation of region shape. They also differ in their ability to distinguish faint features from noise. Differences not previously reported are described, previous discussion where it exits is collated and places where current beliefs may be inaccurate are identified. ESA N91-10646# Oxford Univ. (England). Dept. of Engineering (OUEL-1827/90; ETN-90-97924) Avail: NTIS HC/MF A03 Algorithms for suppressing the effects of image noise on differences for edge finding and image description are addressed. The study proposes that finite differences be computed using a range of operator widths. Although narrow operators are best for describing detailed texture, wide operators report low amplitude responses more reliably. Thus, if wide operators are used to fill gaps in narrow operator responses, each operator can be restricted to report only statistically reliable responses without losing many real features. This sharply reduces the noise in the final output. Theoretical bounds on spurious responses in the finite difference outputs, given only weak assumptions about the signal and noise, are presented. The expected response of the edge finder to an ideal straight step edge is analyzed. These performance measures are compared to those for an algorithm based on smoothing and an algorithm that also considers the spatial structure of noise. Empirical results including measurements of noise in real camera images are presented. N91-10647# Rolls-Royce Ltd., Bristol (England). FULLY AUTOMATIC ANALYSIS IN AN INDUSTRIAL ENVIRONMENT ESA 1. Armstrong and T. Edmunds Derby, England 15 May 1989 12 P Presented at 2nd International Conference on Quality Assurance and Standards in Finite Elements Analysis, May 1989 (PNR-90722; ETN-90-97947) Copyright Avail: NTIS HC/MF A03 Part of the Computer Aided Engineering and Manufacturing (CAEM) project is described. The objective of CAEM is to halve the time between commitment to an engine and entry into service. A two dimensional automatic finite element analysis facility is described. This is designed to contribute to the CAEM project objective by reducing the elapsed time to perform a high quality finite element analysis of the circumferentially averaged response of main engine rotor systems. This allows these analyses to be performed earlier in a design's evolution and increases confidence in the design prior to commitment to manufacture. N91-10359# Meteorological Satellite Center, Tokyo (Japan). IMAGE MAPPING ESA Mitsuo Nesasa In its Meteorological Satellite Center Technical N91-10360# Meteorological Satellite Center, Tokyo (Japan). Masayuki Sasaki In its Meteorological Satellite Center Technical N91-10363# Meteorological Satellite Center, Tokyo (Japan). In its Meteorological Satellite Center Technical Note. Special Issue (1989). Summary of GMS System. 2: Data Processing Mar. 1989 p 75-78 In JAPANESE; ENGLISH N91-10648# Department of Energy, Washington, DC. Robert E. Benner, inventor (to DOE), John L. Gustafson, inventor (DE90-014683; US-Patent-Appl-SN-317389; Patents-US-A7317389) Avail: NTIS HC/MF A04 A parallel computing system and method having improved performance is disclosed. A program is concurrently run on a plurality of nodes for reducing total processing time, each node having a processor, a memory, and a predetermined number of communication channels connected to the node and independently connected directly to other nodes. The present invention improves performance of the parallel computing system by providing a system which can provide efficient communication between the processors and between the system and input and output devices. A method is also disclosed which can locate defective nodes with the computing system. DOE N91-10649# Sandia National Labs., Albuquerque, NM. PERFORMANCE AND OPERATIONAL CHARACTERISTICS OF A FIBER DISTRIBUTED DATA INTERFACE (FDDI) RING Nicholas Testi and Steven A. Gossage 1990 27 p Presented at the 23d NEXUS Conference, Orlando, FL, 11-14 Sep. 1990 (Contract DE-AC04-76DP-00789) (DE90-016103; SAND-90-2066C; CONF-9009232-2) Avail: NTIS HC/MF A03 Sandia National Laboratories is in the process of upgrading the Central Computing Network, which is a large heterogeneous network providing scientific computing, file storage, output services, and remote access to network resources. The migration from the present HYPERchannel-50 technology to HYPERchannel-100 is currently in progress and plans to migrate to the Fiber Distributed Data Interface (FDDI) token ring architecture are being considered. A migration from a variety of proprietary protocols to a primarily TCP/IP environment is also in progress. In order to test the feasibility of the Network Systems Corporation FDDI technology platform, two test rings were constructed. Ring A consists of nine dual attached Data Exchange Units (DXUs) and ring B consists of two dual attached DXUS. The rings are linked together using N715 DXUS. Other DXU models (with associated host computers where applicable) include N130s, an N220, N400s, and FE640 IP routers. Test data on fault isolation and recovery mechanisms, performance, 28 p Workshop held in A. A. Mirin, comp. Aug. 1990 Williamsburg, VA, 25-26 Apr. 1990 (Contract W-7405-eng-48) (DE90-018009; UCRL-JC-104774; CONF-9004240-Summ) Avail: NTIS HC/MF A03 This paper discusses the following topics: Parallel computing at Oak Ridge National Laboratory (ORNL); massive parallelization of fluid plasma turbulence computations; a parallel implementation of the reduced MHD equations; massively parallel Fokker-Planck calculations; massively parallel Fokker-Planck calculations via Monte-Carlo; gyrokinetic calculations on the connection machine; a parallel 2D electrostatic PIC code for the Mark III Hypercube; particle simulation on the NCUBE concurrent multiprocessor; gyrokinetic simulations on massively parallel computers; and, numerical study of collisionless magnetic reconnection on the connection machine. DOE N91-10651# Technische Univ., Delft (Netherlands). Faculty of LABELING PROBLEMS IN TRUTH MAINTENANCE SYSTEMS Along with a new and simple NP-completeness proof of the general labeling problem for Truth Maintenance Systems (TMS), some special cases are discussed and some methods to determine an upperbound for the possible number of labelings for a given TMS are given. Efficient algorithms to estimate this number are used and these algorithms are expected to be used in relabeling TMS networks. ESA N91-10652# Institut National Polytechnique, Grenoble (France). Lab. d'Informatique Fondamentale et d'Intelligence Artificielle. PARALLEL INFERENCE AND COMMUNICATION PROCESS FOR HORN CLAUSE. FIRST ORDER EXTENSION BY CONNECTION METHOD Ph.D. Thesis [INFERENCE PARELLELE ET PROCESSUS COMMUNICANTS POUR LES CLAUSES DE HORN. EXTENSION AU PREMIER ORDRE PAR LA METHODE DE CONNEXION] Maria-Blanca Ibanez-Espiga 1990 253 p In FRENCH; ENGLISH summary Sponsored by CROUS, France and CONICIT, Venezuela (ETN-90-97825) Avail: NTIS HC/MF A12 An OR-parallel inference machine for Horn clause processing is described. This machine is formed by a network of processes ressembling the syntactic structure of the logic program and it uses resolution as its inference mechanism. The FP2 language is used for describing the machine. The communication capabilities of this language are applied for implementing the basic operation in inference: unification. When proving a Horn clause using resolution, the search space contains the axioms of the proof and the current resolvent. This search space is not enough when proving a first order logic formula using Bibel's connection method. At each step, the set of pairs of literals (connections) has to be resolved. This set is called the spanning set of connections. The specification of a correct method for calculating spanning sets of connections is presented. ESA 63 CYBERNETICS Includes feedback and control theory, artificial intelligence, robotics and expert systems. For related information see also 54 Man/System Technology and Life Support. N91-10653# Oak Ridge National Lab., TN. ARTIFICIAL INTELLIGENCE, HUMAN FACTORS, ROBOTICS, (Contract DE-AC05-840R-21400) (DE90-016866; ORNL/FTR-3715) Avail: NTIS HC/MF A03 Traveler was invited to participate in information exchange between Oak Ridge National Laboratory (ORNL) and CISC/Japan Atomic Energy Research Institute (JAERI) on four topics: Artificial Intelligence, Human Factors, robotics, and computer simulation. This exchange took the form of 9 (2-hour) lectures presented by traveler on work done in CS and HF Group, and four presentations by Japanese for traveler's edification. Seven of traveler's lectures were to CISC/JAERI, one to Toshiba Corporation, and one to the Al Steering Committee of JAERI. There was also a presentation by Toshiba Corporation on high frequency work connected with their Boiling Water Reactor (BWR) control room. Final discussion between traveler and JAERI personnel concerned an umbrella agreement with the U.S. Department of Energy (DOE) permitting researcher exchange similar to nuclear researchers. Conclusions are: the U.S. has definite advantages in most areas of Al progress; the Japanese are creating a Monte Carlo radiation dose calculation simulation which will operate at the level of radiating particles (neutrons) with doses calculated for all major organ systems of humans, and major circuits for robots; they are gaining experience in creating major integrated simulations of human/robot activity in a nuclear reactor; and that it would be advantageous for us to have a formal agreement permitting scientists to visit there for more than 15 days at a time. DOE N91-10654# Group. Oxford Univ. (England). Robotics Research IMAGE MOTION EXPERIMENTS WITH THE OXFORD AGV The theory, hardware details, and experiments carried out with transputers controlling the motion of the Oxford AGV (a robotic rover type vehicle), and using Datacube image processing to acquire a sequence of images as the vehicle was moved at different speeds, are described. Techniques to interpret visual motion, including differential camera motion, widely separated views, integrating information along the camera path and techniques not requiring correspondence are reviewed. Basic equations describing optic flow are introduced. Results include an initial view from the vehicle, four translational sequences superimposed on the first frame, and each result shows image features trucked across the sequence when the camera was moved in a straight line. To minimize the effect of rotation the camera should be placed as far back towards the differential point consistent with preserving the required field of view. 64 NUMERICAL ANALYSIS ESA Includes iteration, difference equations, and numerical approximation. N91-10655# Sydney Univ. (Australia). School of Civil and Mining Engineering. SOME RECENT APPLICATIONS OF NUMERICAL METHODS TO GEOTECHNICAL ANALYSIS J. R. Booker, J. P. Carter, J. C. Small, P. T. Brown, and H. G. Poulos May 1989 37 p (PB90-201229; R-598) Avail: NTIS HC/MF_A03 CSCL 12A In geotechnical engineering it is often important to take into account construction path, constitutive modelling and three F. Maffioli and M. T. Galli (AGIP S.p.A., Milan, Italy) 1989 In ITALIAN Sponsored by Italian Ministry of Public The simulated annealing and the Tabu-search heuristic algorithms are evaluated for the solution of a combinatory optimization problem. The experimental results show a slight advantage to the simulated annealing procedure, due to a better use of the additional computing times to obtain increasingly better results. ESA N91-10659# Clarkson Univ., Potsdam, NY. Dept. of Physics. PATH INTEGRAL ASYMPTOTICS IN THE ABSENCE OF CLASSICAL PATHS L. S. Schulman and R. W. Ziolkowski (Lawrence Livermore National Lab., CA.) Jul. 1990 26 p Presented at the 3d Path Integrals from MeV to MeV Conference, Bangkok, Thailand, Jan. 1989 Sponsored in part by NSF (Contract W-7405-eng-48) (DE90-016213; UCRL-CR-104583) Avail: NTIS HC/MF A03 When a boundary value problem in classical mechanics has no solution, the corresponding quantum propagator may be asymptotically dominated by locally flawed classical paths. It is indicated how this assertion can be made precise and it is shown how the propagator is calculated under these circumstances. The essential tools in the arguments will be the use of neutralizers and the path decomposition formula for path integrals. With this method the propagator for barrier penetration is expressed as an N91-10661# Lawrence Livermore National Lab., CA. R. C. Y. Chin and G. W. Hedstrom 1 Aug. 1990 27 p Presented (DE90-016478; UCRL-JC-104693; CONF-9002131-2) Avail: NTIS HC/MF A03 Several examples were given showing the use of asymptotic analysis in the construction of numerical methods based on domain decomposition. Many of the examples come from singularlyperturbed convection-diffusion equations with boundary and internal layers. The final example is less standard in that it deals with the numerical generation of orthogonal polynomials. DOE N91-10662# Los Alamos National Lab., NM. Theoretical Div. STABILITY OF SOME STATIONARY SOLUTIONS FOR THE FORCED Kdv EQUATION Roberto Camassa and T. Yao-Tsu Wu (California Inst. of Tech., Pasadena.) 1990 17 p Presented at the 10th Annual Conference on Nonlinear Science: The Next Decade, Los Alamos, NM, 21-25 May 1990 (Contract W-7405-eng-36) (DE90-016506; LA-UR-90-2744; CONF-9005237-4) Avail: NTIS HC/MF A03 The forced KdV (fKdV) equation has been established by recent studies as a simple mathematical model capable of describing the physics of a shallow layer of fluid subject to external forcing. For a particular one-parameter family of forcings which is characterized by a wave amplitude parameter for supercritical forcing distributions, exact stationary solutions are known. The stability of these solutions was studied as the parameter varies. The linear stability analysis is first carried out, and the structure of the spectrum of the associated eigenvalue problem is discussed using a perturbation approach, about isolated parameter values where eigenfunctions can be expressed in closed form and are the fixed-point solutions of the fKdV equation corresponding to zero eigenvalues. The results identify a set of intervals in the parameter space corresponding to different types of manifestation of instability. In the region of the parameter space where the linear stability analysis fails to provide an answer, a nonlinear analysis was developed to provide a sufficient condition for stability. DOE N91-10663# Los Alamos National Lab., NM. |