摘要:

AbstractIn the range of low defectives, the so called dpm range (<1,000 defectives Eer million), the expensive electrical screening test on electronic components can be replaced by a low cost tightened sampling inspection. Due to the complex electrical test equipment required and the long test sequences applied in the integrated circuit area, especially there the biggest savings in testing costs will be achieved by the introduction of the described tightened sampling inspection method. A certain number of defectives must, however, be tolerated after performance of the tightened sampling inspection, since it is not possible to eliminate all of them by a rectifying sampling test. On the other hand, such inspection undoubtedly ensures the rejection of all scattering lots containing a larger number of defectives. A screening of all rejected lots improves the average outgoing quality in a statistically defined manner to the extent desired by the customer, as described in this paper.If the results achieved by a sampling inspection are to be comparable to those yielded by a screening test, this sampling inspection must of course be very much tighter than it would be necessary for verifying the specified AQL values. The rejection rates up to 20% obtained in the tightened sampling inspection clearly illustrate this statement.For sampling inspection in the dpm range, manufacturers preferably apply single sampling plans with the acceptance number zero. Such sampling plans are very well suited also for the proposed tightened sampling inspection. They lead to simple relations between the sampling parameters and easy-to-read diagrams presented in this paper. These proposed diagrams can be used to select the sampling plan best suited, and to determine quantitatively the average outgoing quality, the average outgoing quality limit and the relative average total inspection after performance of the rectifying tightened sampling inspection. Some numerical examples are given to illustrate the application of the proposed formulas and diagrams.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760085

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractThis paper presents a new parallel conjugate direction method for unconstrained, n-dimensional minimization problems. Starting from one point, the method performs n line searches concurrently along n conjugate directions of the approximate Hessian to derive a new point and generate n new conjugate directions. For deriving the new point, a linear formula is presented, and its convergence properties are discussed. The n new conjugate directions are generated mostly in parallel without explicitly using the approximate Hessian. For quadratic problems, it is shown that the method converges in two parallel iterations. For nonquadratic problems, safeguard steps are taken to guarantee monotonic decency and global convergence. A sufficient condition for“rapid”convergence is also presented. Numerical testing results indicate that, comparing with the conventional conjugate gradient method, the parallel method takes less (parallel) iterations and less function evaluations for all test problems, lmd is most effective for problems whose Hessians change little from iteration to iteration.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760086

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractSingle-and intinite-server queues are diecussed when the arrivals occur at the epochs ot a birth process.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760087

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractIn this paper we show that regular networks possess both max-λand super-λproperties. Given this, the unreliability function for small-edge failure rate can be readily computed.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760088

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractIn the present paper d.c. contactors without economy apparatus re andlyzed. Such devices r quire a careful design of pole shape and an accurate prediction of the dynamical behaviour In order to avoid contact bouncing. Therefore the device analysIs is carried out by a model-ing procedure. In particular, the magnetic field in the dnving electromagnet is evaluated employing a two-dimensional nonlinear finite element technique. The dynamical behaviour of the device is predicted by solving the system of motional and circuIt nonlinear differential equations. The comparison between computational and experimental results shows some discrepan-Clea. Anyway, the procedure allows to satisfactorIly design the pole nhape and to give an approximate estimation ot the critical values of supply voltage.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760089

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractIn this paper we present an overview and critique of the most commonly used techniques for obtaining component reliability estimates as applied toward making series-system reliability predictions. In addition. we present an alternative approach that involves estimating component reliabilityfrom historical life data on previously developed systems. This approach is based upon the well-known theory ofcompeting risks. We present the model and discuss its advantages over other methods. We show how the method of maximum likelihood can be used to obtain parametric estimates of component reliability. In addition we address many of the limitations and assumptions required to ensure its successful application in a“real-world”setting. Thisframework has. in fact, been implemented successfully in a large electronic systems manufacturing firm.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760090

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor

摘要:

AbstractA single bus tightly coupled multiprocessor is a distributed computing system characterized by a single data highway, physically e10se juxtaposition of the processors, and a globally accessible memory that is used for interprocessor communication. We present a prioritized hier rchical Markovian queueing model for such a computer organization. This multiprocessor model consists of a set of processing elemelltg (PEs) and single queue/server pair representing the shared memory. Our model takes into count not only global system behavior, but also behavior of tasks on each processing clement and probabilistic tMk migration among the PEs. We show that the prioritization can be masked by reducing each PE submodel to an equivalent queue/server pair whose scrvice rale depends on the numbcr of tMks in this queue; furthermore, under the Msumption of heavy load, this service rate becomes independent of the number of tasks in the server' queue. 8Med on this reduction, an equivalent multiprocessor model is constructed. Global throughput (both exact and limiting) is then obtained based on this reduction.

ISSN:0228-6203    

DOI:10.1080/02286203.1990.11760091

出版商:Taylor&Francis    

年代:1990

数据来源: Taylor