Thomas R. Bement

Statistical Tolerancing Based on Consumer's Risk Considerations

In a production process it is necessary to specify tolerances for characteristics, such as dimensions within which the measured characteristic must fall in order for the part to be acceptable. Such tolerances may often be set informally, but a considera..

On a Method for Detecting Clusters of Possible Uranium Deposits

When a two-dimensional map contains points that appear to be scattered somewhat at random, a question that often arises is whether groups of points that appear to cluster are merely exhibiting ordinary behavior that one can expect with any random distribution of points, or whether the clusters are too pronounced to be attributable to chance alone. In this paper a method for detecting clusters along a straight line.is applied to the two-dimensional map of bismuth-214 anomalies observed as part of the National Uranium Resource Evaluation Program in the Lubbock, Texas region. Some exact probabilities associated with this method are computed and compared with two approximate methods. The two methods for approximating probabilities work well in the cases examined, and can be used when it is prohibitive to obtain the exact probabilities.

Locating maximum variance segments in sequential data

An automated method is presented for the identification of peaks in sets of sequential data. The method is based upon the location of those segments with maximum variance and has the advantage of guarding against the masking of small-scale effects by large-scale effects. The procedure is illustrated with data taken as part of the National Uranium Resource Evaluation project.

Ch. 11. Predict: A new approach to product development and lifetime assessment using information integration technology

Publisher Summary Performance and reliability evaluation with diverse information combination and tracking (PREDICT) is a successful example of information integration technology that has been applied in two parallel applications: automotive system development and stockpile physics packages in nuclear weapons. This chapter outlines the applications, implementation steps, expert judgment, statistical tools, and decision making that make up the PREDICT methodology. PREDICT has demonstrated its effectiveness for expertise capture, reliability, and performance estimation in the nuclear weapons program and for concept system development in the automotive industry. In the post–Cold War era, the basic philosophy of information integration has been positively affecting the certification process of the nuclear systems. This same philosophy has been providing the formal structure for taking advantage of a companys greatest asset—the knowledge and expertise of its engineers and designers.

Characterizing reliability in a product/process design-assurance program

Just as estimates of cost and program timing are critical factors to be known and monitored during a new product development program, so too is the reliability perspective. This paper describes an approach to reliability modeling that encompasses the impact of both product and manufacturing process design on the distribution (characterizing the uncertainty) of reliability over time. It further describes the elicitation of expert judgment which is used to quantify the initial reliability estimate, including uncertainty. Finally, it describes a Bayesian updating approach which is applicable throughout the development program, and which accommodates a wide variety of possible new information. Although the model is rigorous in its execution; a user-friendly approximation is also described which may be useful to the product development team for purposes of test and validation planning.

Discriminant analysis applied to aerial radiometric data and its application to Uranium favorability in South Texas

In an effort to establish radiometric signatures of geologic units, 10 discriminant analysis techniques were applied to aerial radiometric data collected along the Texas Gulf Coast for the Grand Junction, Colorado, Office, U.S. Department of Energys National Uranium Resource Evaluation (NURE)program. Results of this study show that partial discriminant analysis with the linear discriminant function (LDF)applied to the raw data is useful for establishing radiometric signatures and for classifying new observations. Signatures for favorable and unfavorable units along the Texas Gulf Coast were established and new observations were then classified as favorable or unfavorable for hosting uranium deposits based on the established training sets.

Estimating percentiles in aerial radiometric data using normal and lognormal distributional assumptions

Implicitly or explicitly, percentile estimation is an important aspect of the analysis of aerial radiometric survey data. Standard deviation maps are produced for quadrangles which are surveyed as part of the United States Department of Energys National Uranium Resource Evaluation. These maps show where variables differ from their mean values by more than one, two, or three standard deviations. Data may or may not be log-transformed prior to analysis. These maps have specific percentile interpretations only when proper distributional assumptions are met. Monte Carlo results are presented in this paper which show the consequences of estimating percentiles by (1)assuming normality when the data are really from a lognormal distribution, and (2)assuming lognormality when the data are really from a normal distribution.

Correlation of the Cesium-134/Cesium-137 Ratio to Fast Reactor Burnup

The correlation of 134Cs/137Cs to burnup in irradiated fast reactor fuel pins was investigated using nondestructive precision gamma-ray scanning. This correlation was significant within individual subassemblies provided certain basic assumptions are satisfied. The calibration line must be calculated for each assembly based on independently measured burnup values. Differences in calibration lines for separate subassemblies are highly significant (P < 0.005). Therefore, application of this technique for the nondestructive measurement of burnup in fast reactor fuels has serious problems that must be considered.

Recovery of Interblock Information for Designs with Groups of Correlated Blocks

Abstract A procedure is presented for estimating the dispersion matrix of block totals in an incomplete block design where blocks can be divided into groups such that those from different groups are independent and those from the same group are dependent. Equal block variances and equal nonzero covariances are assumed. It is shown that when the estimated dispersion matrix is used to obtain a weighted least-squares interblock estimator of treatment means, both that estimator and the combined intra-interblock estimator converge in mean square to their respective BLUEs more rapidly than the BLUEs converge to the vector of treatment means.

Identification of regions enriched or depleted in radioelements through nondistributional analysis of aerial radiometric data

Abstract Throughout the aerial radiometric reconnaissance survey portion of the U.S. Department of Energys National Uranium Resource Evaluation (NURE) program, the identification of outliers (anomalies) was an important approach to locating regions with radio-element concentrations that are either higher or lower than expected. The method introduced herein to locate such regions involves three steps: selection of a high (or low) threshold for the variate of interest; use of the sample percentile to identify all points of interest; and movement of a window over the selected data to locate significant clusters of observations. These steps, applied to aerial radiometric 214 Bi (equivalent uranium) data collected over the Copper Mountain area in Wyoming, resulted in the identification of areas enriched in that variate.