206
World Abstracts on Microelectronics and Reliability
selection of efficient testing strategies for reparable systems comprised of components arranged in series. Two cost models (for perfect and imperfect testing) represent the consequences of possible test realizations. The probability that any particular component is responsible for the failure is derived and used as a basis for the two models. The model for perfect testing is solved exactly. In the optimal perfecttest sequence the components are tested in decreasing order of the ratio of: [probability that the component is responsible for the system failure] to [component test cost]. For imperfect testing, possible diagnostic errors are included in a model for which two heuristic solution strategies are provided. The model represents the consequences of both false-positive and false-negative component-test outcomes. The heuristic strategies yield efficient test sequences. Under reasonable assumptions, the second heuristic strategy is guaranteed to locate the optimal test sequence. The model can evaluate quantitatively the benefits of test-accuracy enhancement plans. The utility of the models and algorithms is that they provide convenient methods for selecting efficient testsequences. This is illustrated by representative examples.
Using test data to predict avionics integrity. GLEN E. BENZ. Proc. A. Reliab. Maintainab. Syrup., 1 (1990). A producer may want to use his in-house test data to predict avionics integrity of a new product because these data provide: (I) ages at failure, (2) measures of manufacturing integrity and (3), insights into reasonable acceleration schedules for the Durability/Economic Life Test. Test observations of ages to failure and end of test are adjusted by a single factor that represents both stress and complexity differences between new product usage and historical product testing. These adjusted age observations are then ordered and fitted with a straight line probability law. The producer can show that his historical failure free operating period, as applied to the new product, will meet the requirement with some degree of confidence. An analytic approach to performing a maintainability demonstration. ELDONC. GOULDEN. IEEE Trans. Reliab. 39(1), 19 (1990). US Mil-Std-471A defines a planned program to demonstrate maintainability at the equipment level. This paper develops a method to demonstrate maintainability at both the equipment and system levels. Empirical data are used as a basis for analytic extrapolation. Both manual and automatic fault-localization are used. Several other concepts are defined and then explained by examples. Two major benefits from this method are: (1) a better statistical estimate of the mean time-to-repair of an equipment or system, and (2) the savings in time and money by analytically estimating repair actions rather than physically demonstrating them. An appendix explains the stratification of faults. Error analysis for optimal design of accelerated tests. D. J. HANNAMAN, N. ZAMANI, J. DHIMAN and M. G, BUEHLER. 28th /1. Proc. Reliab. Phys. Syrup. (IEEE), 55 (March 1990). In accelerated life testing the operating life of a device is determined by extrapolation from the failure data at a set of stress points. A methodology has been developed for selecting an optimal set of stress points given a constraint in available test time. The approach is based on linearizing the failure equation and applying multiple linear regression to obtain the optimal stress points. Specific application is made to Black's model of electromigration. A time-dependent availability measure for a reparable system subject to catastrophic failures. G. K. AGRAFIOTIS. IEEE Trans. Reliab. 39(l), 23 (1990). This paper deals with a reparable system which is subject to catastrophic failures while it is in operation or under repair. The topic of investigation is the time-dependent point availability measure of the system which is obtained in terms of the
repair and failure time distributions and in the form of convolution integrals which are readily evaluated by means of known computational algorithms. Also derived are expressions for this measure which are useful for estimating the parameters of the model according to data available. An application of the model is considered.
Classical and Bayes approaches to environmental stress screening (ESS): a comparison. RICHARD E. BARLOW, IGOR BAZOVSKYSR and SERGIOWECHSLER. Proc./t. Re[lab. Maintainab, Syrup., 81 (1990). Optimal designs for Environmental Stress Screening (ESS) plans are discussed relative to both a classical and a Bayesian statistical point of view. A solution to the problem of determining optimal screen time durations given the stress level as well as the optimal stress level given a screen duration is given using the Bayesian approach. Other ESS measures described in a recent Military Handbook are calculated using our Bayesian approach. Use of advanced analytical techniques for VLSI failure analysis. INDRAJIT BANERJEE, BRYAN TRACY, PAUL DAVIES and BOB McDONALD. 28th A. Proc. Reliab. Phys. Syrup. (IEEE)~ 61 (March 1990). Advanced analytical tools that have been primarily used for research and development application in the past are now being used extensively to understand VLSI failure analysis and reliability problems. In this paper we present several new applications and techniques in the use of scanning electron microscopy (SEM), transmission electron microscopy (TEM), focused ion beam (FIB) microsurgery and secondary ion mass spectroscopy (SIMS) for problem solving. These tools, used in new ways, are playing a key role in identifying the sources of defects leading to potential device reliability problems and contributing to the elimination of their sources in the process line. Pi-faetors revisited. RAYMOND H. SEIDL and WILL1AMJ. GARRY. Proc. A. Reliab. Maintainab. Syrup., 19 (1990). As part of an RADC contract to develop new failure rate models in MIL-HDBK-217E for microcircuits, the need for updated values of three failure rate adjustment factors ("pi-factors") became evident. These pi-factors are nQ (the quality factor), n~ (the environmental factor), and ~L (the learning factor). The logic, methodology, and results of the approach taken in developing these factors are addressed, and a discussion on how to use each pi-factor is provided. Importance and sensitivity analysis in assessing system reliability. A. GANDINI. IEEE Trans. Reliab. 39(1), 61 (1990). After reviewing various importance concepts adopted in reliability, a method for sensitivity analysis is proposed. The method uses the heuristically based generalized perturbation theory (GPT) methodology, widely adopted in reactorphysics studies. The concept of importance of a state in the Markov model representation of systems is introduced. The resulting formulations apply to any response of interest in reliability analysis. The relationship between the GPT method and Birnbaum importance is also given. Expert systems maintainability, CHRISTINEW. J. CHEE and MARGARETA. POWER. Proc. A. Reliab. Maintainab, Symp., 415 (1990). Because expert systems are a new software technology, developers have given little attention to the problem of expert system maintenance. This paper presents an expert system development process tailored for the maintainability design support role, and discusses related issues. These issues include the software engineering process, the differences between expert systems and conventional software which affect maintainability, and the use of expert system shells and development tools.