Publications

The objective of this reseach is the investigation of alternative methods for characterizing the reliability of systems with time dependent failure modes associated with stockpile aging. Reference to 'reliability degradation' has, unfortunately, come to be associated with all types of aging analyes: both deterministic and stochastic. In this research, in keeping with the true theoretical definition, reliability is defined as a probabilistic description of system performance as a funtion of time. Traditional reliability methods used to characterize stockpile reliability depend on the collection of a large number of samples or observations. Clearly, after the experiments have been performed and the data has been collected, critical performance problems can be identified. A Major goal of this research is to identify existing methods and/or develop new mathematical techniques and computer analysis tools to anticipate stockpile problems before they become critical issues. One of the most popular methods for characterizing the reliability of components, particularly electronic components, assumes that failures occur in a completely random fashion, i.e. uniformly across time. This method is based primarily on the use of constant failure rates for the various elements that constitute the weapon system, i.e. the systems do not degrade while in storage. Experience has shown that predictions based upon this approach should be regarded with great skepticism since the relationship between the life predicted and the observed life has been difficult to validate. In addition to this fundamental problem, the approach does not recognize that there are time dependent material properties and variations associated with the manufacturing process and the operational environment. To appreciate the uncertainties in predicting system reliability a number of alternative methods are explored in this report. All of the methods are very different from those currently used to assess stockpile reliability, but have been used extensively in various forms outside Sandia National Laboratories. It is hoped that this report will encourage the use of 'nontraditional' reliabilty and uncertainty techniques in gaining insight into stockpile reliability issues.

This paper outlines the results of a cooperative effort between Sandia National Laboratories, Associated Power Analysts, Inc. and Texas A&M University to characterize the impact of a changing regulatory environment on the reliability of customer electrical service. It was desired to assess the impact in as realistic an environment as possible. Due the availability of data the initial study centered on the electric power grid in Texas. Specifically, data from the Electric Reliability Council of Texas (ERCOT) for the 1997 operational year was used in the research. Based on geography and location of generation and transmission lines, ten basic areas were considered and each area was modeled as a single point generation and load. A number of restructuring scenarios were developed by researchers at Sandia National Laboratories and investigated by Associated Power Analysts using their N-Area Reliability Program (NARP). The present study is limited to an assessment of the adequacy aspects of reliability: sufficiency of installed generation and transmission capacity to satisfy the needs of all consumers in a steady- state sense. The results are, on one hand conservative in that they address on] y the impact of peak loading. Alternatively, the~ are optimistic in that the transmission lines are assumed to be in continuous operation. The major results of this study indicate that, in anew regulatory era, the reliability of customer service will be significantly impacted, possibly in a negative fashion, unless the effects of the ensuing economic pressures are understood and appropriate actions taken.

This report provides an introduction to the various probabilistic methods developed roughly between 1956--1985 for performing reliability or probabilistic uncertainty analysis on complex systems. This exposition does not include the traditional reliability methods (e.g. parallel-series systems, etc.) that might be found in the many reliability texts and reference materials (e.g. and 1977). Rather, the report centers on the relatively new, and certainly less well known across the engineering community, analytical techniques. Discussion of the analytical methods has been broken into two reports. This particular report is limited to those methods developed between 1956--1985. While a bit dated, methods described in the later portions of this report still dominate the literature and provide a necessary technical foundation for more current research. A second report (Analytical Techniques 2) addresses methods developed since 1985. The flow of this report roughly follows the historical development of the various methods so each new technique builds on the discussion of strengths and weaknesses of previous techniques. To facilitate the understanding of the various methods discussed, a simple 2-dimensional problem is used throughout the report. The problem is used for discussion purposes only; conclusions regarding the applicability and efficiency of particular methods are based on secondary analyses and a number of years of experience by the author. This document should be considered a living document in the sense that as new methods or variations of existing methods are developed, the document and references will be updated to reflect the current state of the literature as much as possible. For those scientists and engineers already familiar with these methods, the discussion will at times become rather obvious. However, the goal of this effort is to provide a common basis for future discussions and, as such, will hopefully be useful to those more intimate with probabilistic analysis and design techniques. There are clearly alternative methods of dealing with uncertainty (e.g. fuzzy set theory, possibility theory), but this discussion will be limited to those methods based on probability theory.

The design of complex systems is difficult at best, but as a design becomes intensively dependent on the computer processing of external and internal information, the design process quickly borders chaos. This situation is exacerbated with the requirement that these systems operate with a minimal quantity of information, generally corrupted by noise, regarding the current state of the system. Establishing performance requirements for such systems is particularly difficult. This paper briefly sketches a general systems design approach with emphasis on the design of computer based decision processing systems subject to parameter and environmental variation. The approach will be demonstrated with application to an on-board diagnostic (OBD) system for automotive emissions systems now mandated by the state of California and the Federal Clean Air Act. The emphasis is on developing approach for establishing probabilistically based performance requirements for computer based systems.

This paper outlines the use of a failure modes and effects analysis for the safety assessment of a robotic system being developed at Sandia National Laboratories. The robotic system, the weigh and leak check system, is to replace a manual process for weight and leakage of nuclear materials at the DOE Pantex facility. Failure modes and effects analyses were completed for the robotics process to ensure that safety goals for the systems have been met. Due to the flexible nature of the robot configuration, traditional failure modes and effects analysis (FMEA) were not applicable. In addition, the primary focus of safety assessments of robotics systems has been the protection of personnel in the immediate area. In this application, the safety analysis must account for the sensitivities of the payload as well as traditional issues. A unique variation on the classical FMEA was developed that permits an organized and quite effective tool to be used to assure that safety was adequately considered during the development of the robotic system. The fundamental aspects of the approach are outlined in the paper.

No abstract available.

Emphasis on the human-to-aircraft interface has magnified in importance as the performance envelope of today`s aircraft has continued to expand. A major problem is that there has been a corresponding increase in the need for better fitting protection equipment and unfortunately it has become increasingly difficult for aircrew members to find equipment that will provide this level of fit. While protection equipment has, historically had poor fit characteristics, the issue has grown tremendously with the recent increase in the numbers of minorities and women. Fundamental to this problem are the archaic methods for sizing individual equipment and the methods for establishing a sizing system. This paper documents recent investigations by the author into developing new methods to overcome these problems. Research centered on the development of a new statistically based method for describing form and the application of fuzzy clustering using the new shape descriptors. A sizing system was developed from the application of the research, prototype masks were constructed and the hardware tested under flight conditions.