Publications

We present an algorithm for the QR factorization of a dense matrix without column pivoting on a hypercube multiprocessor. The algorithm combines the optimal numerical efficiency of Householder reflections with the excellent communication properties of the torus wrap mapping. Analytical results indicate that the communication cost for this algorithm is less than that for other common approaches. Numerical results on an nCUBE 2 confirm the efficiency of our technique. 23 refs., 5 figs., 1 tab.

The spherical element computer code DMC (Distinct Motion Code) used to model rock motion resulting from blasting has been enhanced to allow routine computer simulations of bench blasting. The enhancements required for bench blast simulation include: (1) modifying the gas flow portion of DMC, (2) adding a new explosive gas equation of state capability, (3) modifying the porosity calculation, and (4) accounting for blastwell spacing parallel to the face. A parametric study performed with DMC shows logical variation of the face velocity as burden, spacing, blastwell diameter and explosive type are varied. These additions represent a significant advance in the capability of DMC which will not only aid in understanding the physics involved in blasting but will also become a blast design tool. 8 refs., 7 figs., 1 tab.

A wide variety of physical phenomena arising within many scientific disciplines can be described by systems of coupled partial differential equations (PDEs). The numerical approximation of these PDEs often involves the solution of a system of algebraic equations (possibly nonlinear) which are typically large, sparse and nonsymmetric. The increasing computational demands required by the solution of such complex scientific applications has motivated the current direction toward large-scale parallel computers. We, therefore, consider solution techniques of representative systems of equations on large scale MIMD machines. Our primary emphasis in this study is the evaluation of iterative methods for the solution of nonsymmetric systems. In particular, we discuss two Krylov subspace methods, the conjugate gradient squared algorithm (CGS) and the generalized minimum residual method (GMRES), along with the multigrid algorithm (MG) on massively parallel MIMD architectures. The focus of this evaluation considers the performance of various algorithm and implementation variations over a broad selection of problems using a parallel machine.

A method which is capable of efficient calculation of the axisymmetric flow field produced by a large system of ring vortices is presented in this report. The system of ring vortices can, in turn, be used to model body surfaces and wakes in incompressible unsteady axisymmetric flow fields. This method takes advantage of source point and field point series expansions which enables one to make calculations for interactions between groups of vortices which are in well separated spatial domains rather than having to consider interactions between every pair of vortices. In this work, series expansions for the stream function of the ring vortex system are obtained. Such expansions explicitly contain the radial and axial velocity components. A Fortran computer code RSOLV has been written to execute the fast solution technique to calculate the stream function and the axial and radial velocity components at points in the flow field. Test cases have been run to optimize the code and to benchmark the truncation errors and CPU time savings associated with the method. Non-dimensional truncation errors for the stream function and total velocity field are on the order of 5 {times} 10{sup {minus}5} and 3 {times} 10{sup {minus}3} respectively. Single precision accuracy produces errors in these quantities up to about 1 {times} 10{sup {minus}5}. For 100 vortices in the field, there is virtually no CPU time savings with the fast solver. For 10,000 vortices in the flow, the fast solver obtains solutions in about 1% to 3% of the time required for the direct solution technique. Simulations of vortices with square and circular cores were run in order to obtain expressions for the self-induced velocities of such vortices. 8 refs., 26 figs.

Parachute system performance issues such a turnover and wake recontact may be strongly influenced by velocities induced by the wake of the delivering aircraft, especially if the aircraft is maneuvering at the time of parachute deployment. The effect of the aircraft on the parachute system is a function of the aircraft size, weight, and flight path. In order to provide experimental data for validation of a computer code to predict aircraft wake velocities, a test was conducted in the NASA 14 {times} 22 ft wind tunnel using a 5.78% model of the B-1B strategic bomber. The model was strut mounted through the top of its fuselage by a mechanism which was capable of pitching the model at moderate rates. In this series of tests, the aircraft was pitched at 10{degree}/sec from a cruise angle of attack of 5.3{degree} to an angle of attack of 11{degree} in order to simulate a 2.2g pullup. Data were also taken for the subsequent pitch down sequence back to the cruise angle of attack. Instantaneous streamwise and vertical velocities were measured in the wake at a number of points using a hot wire anemometer. These data have been reduced to the form of downwash coefficients which are a function of the aircraft angle of attack time-history. Unsteady effects are accounted for by use of a wake convection lag-time correlation. 12 refs., 59 figs., 4 tabs.

Archimedes is a prototype mechanical assembly system which generates and executes robot assembly programs from a CAD model input. The system addresses the unrealized potential for flexibility in robotic mechanical assembly applications by automating the programming task. Input is a solid model of the finished assembly. Parts relationships and geometric constraints are deduced from the solid model. A rule-based planner generates a generic'' assembly plan that satisfies the geometric constraints, as well as other constraints embodied in the rules. A retargetable plan compiler converts the generic plan into code specific to an application environment. Execution of the compiled plan in a workcell containing an Adept Two robot, a vision system, and other parts handling equipment will be shown on videotape.

Recent experiments on Lawrence Livermore's ATA indicate that there may be problems with the IFR (Ion Focused Regime) transport of the relativistic electron beam (REB) through the ATA accelerator. For beam currents greater than about 7-kA, the beam is observed to be inverse-tailored with the beam radius increasing from beam head to tail. This inverse-tailoring is considered unfavorable for endo- atmospheric beam propagation. An unusual feature of ATA's laser- produced IFR channel is that it has a rectangular cross-section. One possible explanation for the lack of good beam transport may be ion motion in the IFR channel which disrupts the tail of the beam. In this report an ATA-like electron beam, propagating on a laser-ionized rectangular IFR channel is simulated using the 3-D magnetostatic code BUCKSHOT. The simulations demonstrate that non-axisymmetric ion motion, similar to the ion hose instability, can produce an inverse- tailored electron beam similar to those found in the experiment. The simulation results should be relevant to upcoming TROLL experiments with a laser-ionized IFR channel, with several beam parameters similar to ATA's. 1 ref., 22 figs.

One critical issue to be addressed in the compact recirculating linac program concerns optimal beam injection into a racetrack-shaped accelerator. There are at least three candidates, axial beam injection, tangential beam injection, and laser-channel-assisted beam injection. In this report these three approaches are examined using computer simulation techniques. 3 refs., 27 figs., 2 tabs.

A very simple algorithm is presented that allows particles to be loaded or initialized in a particle simulation code. The algorithm can load particle positions or velocities according to any well- behaved density or distribution function. Sample codes are given in 1, 2, and 3 dimensions. The technique is highly efficient. 2 refs., 9 figs., 3 tabs.

Preliminary BUCKSHOT simulations of a recirculating linear accelerator have been made. Two accelerator configurations have been examined for a variety of beam currents (10-40 kA). The first configuration is an attempt to simulate conditions accessible to near-term experiments. The second configuration tries to mock up a next generation application type machine. 3 figs.

A method which is capable of calculating the unsteady flow field around circular-arc bluff bodies of zero thickness is presented. This method utilizes linear vortex panels to model the body surface and a portion of the wake surfaces. Discrete vortices are used to model the remainder of the wake surfaces. Separation is assumed to occur at the sharp edges of the bodies. Numerical results for circular-arc bodies with included angles of less than 180/degree/ are compared with experimental data and found to be in good agreement. 31 refs., 15 figs.