Accelerated Degradation Tests: Modeling and Analysis

High reliability systems generally require individual system components having extremely high reliability over long periods of time. Short product development times require reliability tests to be conducted with severe time constraints. Frequently few or no failures occur during such tests, even with acceleration. Thus, it is difficult to assess reliability with traditional life tests that record only failure times. For some components, degradation measures can be taken over time. A relationship between component failure and amount of degradation makes it possible to use degradation models and data to make inferences and predictions about a failure-time distribution. This article describes degradation reliability models that correspond to physical-failure mechanisms. We explain the connection between degradation reliability models and failure-time reliability models. Acceleration is modeled by having an acceleration model that describes the effect that temperature (or another accelerating variable) has on the rate of a failure-causing chemical reaction. Approximate maximum likelihood estimation is used to estimate model parameters from the underlying mixed-effects nonlinear regression model. Simulation-based methods are used to compute confidence intervals for quantities of interest (e.g., failure probabilities). Finally we use a numerical example to compare the results of accelerated degradation analysis and traditional accelerated life-test failure-time analysis.     

Bootstrap methods for statistical inference from stereological estimates of volume fraction

We suggest the use of bootstrap methods for inference from stereological estimates of volume fraction. An informal introduction to stereological estimation of volume fraction and to principles of bootstrap techniques is given. The bootstrap method is a robust computer-intensive resampling technique, based on independent random sampling from a data set with replacement. Bootstrap methods were used to estimate confidence intervals for volume fractions, and to test for a significant difference between estimated volume fractions from two samples. Two sampling designs are considered: independent replicated samples (visual fields) from a single object, and estimates of volume fraction from multiple independent objects. The methods are presented as worked examples on real data sets obtained from tumour pathology (mammary cancer, pancreatic cancer). The volume fraction of glandular lumina per total volume of the epithelial phase was chosen as target parameter. It indicates the degree of glandular differentiation in adenocarcinomas and is estimated as a ratio-of-means statistic with variable denominator within cases. The confidence intervals of the volume fraction estimated by the bootstrap method were slightly narrower than the parametrically calculated confidence intervals for all data sets. The outcomes of significance tests based on the bootstrap technique were unchanged as compared with classical tests based on the assumptions of normality and homoscedasticity of the data. Special attention was paid to the reproducibility of the bootstrap technique in replicated trials on the same data.     

Boosting Bootstrap FLD Subspaces in Classification

Propose a new ensemble learning algo- rithm based on Fisher linear discriminant analysis （FLD or LDA） and Adaboost.M2 for multi-class classification problem. On each iteration of boosting, a number of bootstrap training subsets are first sampled from the original training set according to the sample weights, and FLD is employed in each subset to construct a number of bootstrap FLD subspaces. Then the subspace with the minimum weighted K- nearest neighbor （KNN） pseudo-loss was selected to form the final strong classifier and update the sample weights. The main advantage of our algorithm is that its application is not restricted to the Gaussian assumption of data distribution, and the learning criterion is directly related to the minimum training error. Through the experiments on two real-world data sets, we demonstrate that our algorithm is superior to other techniques, especially for complex data distribution.     

Regularized linear and kernel redundancy analysis

Redundancy analysis (RA) is a versatile technique used to predict multivariate criterion variables from multivariate predictor variables. The reduced-rank feature of RA captures redundant information in the criterion variables in a most parsimonious way. A ridge type of regularization was introduced in RA to deal with the multicollinearity problem among the predictor variables. The regularized linear RA was extended to nonlinear RA using a kernel method to enhance the predictability. The usefulness of the proposed procedures was demonstrated by a Monte Carlo study and through the analysis of two real data sets.     

Correlation analysis of principal components from two populations

We investigate a correlation coefficient of principal components from two sets of variables. Using perturbation expansion, we get a limiting distribution of the correlation. In addition, we obtain a limiting distribution of the Fisher's z transformation of the above correlation. Additionally, we verify the accuracy of the limiting distributions using Monte Carlo simulations. Finally in this study, we present two examples and a bootstrap estimation.     

A bootstrap approach to visual user-interface design and development

An approach to the design of effective computer-based systems is discussed. This approach exploits the user's traditional diagrammatic notations in an effort to achieve usability for experts other than computer professionals. Notations are formalized as visual languages, thus allowing the design of visual editors, interpreters, and compilers. The users themselves exploit these tools to define a hierarchy of environments by a bootstrapping approach. By navigating within these environments, they can progressively design visual interfaces and computing tools that allow them not only to execute the required computational tasks, but also to gain insight into and control the computational process, and check the results.     

Computational statistics for cell survival curves. II. Evaluation of colony-forming ability of a group of cell strains by the APL function CFAGROUP

An interactive program for the statistical analysis of groups of cell survival curves is described. Methods from linear regression are applied in the presence of repeated experiments and estimates of a common slope and its variance are obtained by weighted means and a bootstrap resampling technique.     

Estimation of Life with Piecewise Linear-Quadratic Hazard Rate

Bladder is an important accessory in the curing operation for tire manufacturing. Failure of bladder, during its use, results in scrap of a large amount of tire under curing. This leads to heavy financial loss to the company. In this article, we deal with estimation of life and determination of replacement plan for bladder. The life of bladder has been modeled using a piecewise linear-quadratic hazard function. A computational procedure is proposed for estimation of life. Finally, a replacement plan for bladder is derived.     

A LIS compiler for GCOS-7

This paper describes the implementation of a LIS compiler for GCOS-7. LIS is a high level system implementation language developed at CII-Honeywell Bull during the middle 1970s, and experience with the language and its implementation have largely influenced the design of Ada. The design of the compiler was particularly aimed at efficient code generation. Design decisions concerning the run-time organization in relation to procedure call and separate compilation are discussed. The structure of the compiler is described. The articulation between the different phases of the code generator is emphasized. Experience with the bootstrap is related.     

Bootstrap learning for accurate onset detection

Supervised learning models have been applied to create good onset detection systems for musical audio signals. However, this always requires a large set of labeled training examples, and hand-labeling is quite tedious and time consuming. In this paper, we present a bootstrap learning approach to train an accurate note onset detection model. Audio alignment techniques are first used to find the correspondence between a symbolic music representation (such as MIDI data) and an acoustic recording. This alignment provides an initial estimate of note boundaries which can be used to train an onset detector. Once trained, the detector can be used to refine the initial set of note boundaries and training can be repeated. This iterative training process eliminates the need for hand-labeled audio. Tests show that this training method can improve an onset detector initially trained on synthetic data. Major part of work was done while the first author was at Carnegie Mellon University. Editor: Gerhard Widmer