ENBIS-8 in Athens

21 – 25 September 2008 Abstract submission: 14 March – 11 August 2008

My abstracts

 

The following abstracts have been accepted for this event:

  • Interpreting Multivariate Control Chart Signals using Computational Intelligence Techniques

    Authors: V.P. Plagianakos Department of Informatics with Biomedical Applications, University of Central Greece, Papasiopoulou 2-4, Lamia, 35100, Greece. S. Bersimis Department of Informatics & Telematics, Harokopio University, 89, Harokopou Street, 176 71, Kallithea, Greece.
    Primary area of focus / application:
    Submitted at 17-May-2008 19:28 by Vassilis Plagianakos
    Accepted
    22-Sep-2008 16:00 Interpreting Multivariate Control Chart Signals using Computational Intelligence Techniques
    Multivariate control charts are used for monitoring and controlling the process mean and the process variability of multivariate processes. These control charts are able to recognize an out-of-control process. The identification of an out-of-control variable or variables after a multivariate control chart signals has been an interesting topic for many researchers over the last few years. This work reviews promising techniques for interpreting an out-of-control signal, while introduces new algorithms and models based on recently proposed computational intelligence techniques. The proposed methods, which have a sound mathematical background, are thoroughly investigated and have proven to be efficient and effective.
  • Interpreting Shewhart Type Control Chart Signals Using Pattern Based Rules and Artificial Intelligence Techniques

    Authors: V.P. Plagianakos Department of Informatics with Biomedical Applications, University of Central Greece, Papasiopoulou 2-4, Lamia, 35100, Greece. A. Kallioras National Statistical Service of Greece, Kiprou 38, Lamia, 35100, Greece S. Bersimis Department of Informatics & Telematics, Harokopio University, 89, Harokopou Street, 176 71, Kallithea, Greece.
    Primary area of focus / application:
    Submitted at 17-May-2008 20:57 by Athan Kallioras
    Accepted
    23-Sep-2008 12:20 Interpreting Shewhart Type Control Chart Signals Using Pattern Based Rules and Artificial Intelligence Techniques
    Shewhart type control charts are widely used in Phase I for both, making a decision about the status of the process and estimating the parameters of the process. Specifically, Shewhart type control charts indicate whether the process under study is in-control or it exhibits a non-random behavior. Subsequently, if the process is under statistical control, Shewhart type control charts can be used to estimate the process parameters. Alternatively, if the process exhibits a non-random behavior, Shewhart type control charts can be used for identifying which is the exact problem of the process. In this work, two new procedures are proposed that can improve the ability of Shewhart type control charts to identify the exact problem of the process in Phase I. These procedures can be applied to control charts for the mean and control charts for the dispersion of the process. The first method uses pattern based rules while the second one uses computational intelligence techniques. These two procedures are compared against alternative approaches giving very interesting results.
  • Fast Semiparametric Models for Active Network Tomography

    Authors: George Michailidis
    Affiliation: The University of Michigan
    Primary area of focus / application:
    Submitted at 2-Jun-2008 08:30 by David Steinberg
    Accepted
    23-Sep-2008 09:40 Fast Semiparametric Models for Active Network Tomography
    In this talk we discuss
    the problem of active network tomography in which probe traffic is injected across a network.
    The end-to-end statistics are collected and used to infer the link-level statistics. We consider the use
    of simple, yet flexible continuous models that are appropriate for monitoring scenarios and are an
    improvement over previously consider discretized delay models. In addition, we develop a moment
    estimation scheme that allows for fast, computationally
    efficient implementation with the goal of monitoring. The performance of the
    proposed methodology is exhibited on a number of simulated and emulated network scenarios
  • Dynamic Design of Large-Scale Validation Experiments

    Authors: Nikolaus Haselgruber, Hannes Hick
    Affiliation: AVL List GmbH
    Primary area of focus / application:
    Keywords: Design of Experiments, Automotive Industry
    Submitted at 16-Jun-2008 08:56 by Nikolaus Haselgruber
    Accepted (view paper)
    23-Sep-2008 11:20 Dynamic Design of Large-Scale Validation Experiments
    In the automotive industry, an important field of experimentation is system validation. The aim is to demonstrate that the system (e.g. engine, transmission, exhaust gas aftertreatment system, complete vehicle, etc.) fulfills the technical specifications (e.g. in terms of a reliability target) as well as the customer expectations. The system is required to be of sufficient maturity at the beginning of the validation phase so that there is a realistic chance to pass the validation program without failures. Consequently, the objective of these type of experimentation is not to optimize a product but to optimize its validation program. This is done by a systematic and representative variation of system load parameters spanning the customer usage space.



    The presentation shows an algorithm including examples of this DoE adaptation which has been applied successfully in different vehicle development projects.
  • A Heuristic Method for Marketing Surveys

    Authors: Stefano Barone, Alberto Lombardo and Pietro Tarantino
    Affiliation: University of Palermo, Department of Manufacturing Technology and Managerial Engineering
    Primary area of focus / application:
    Keywords: Marketing, Surveys, preference capturing
    Submitted at 25-Jul-2008 11:23 by Stefano Barone
    Accepted
    22-Sep-2008 11:39 A Heuristic Method for Marketing Surveys
    Many decision-making theories and models derived from the integration of knowledge from several fields, such as cognitive and motivational science, psychology, psychometrics, communication and information science, sociology and statistics. These theories are particularly used for the identification of important attributes composing the product and the service under study. The knowledge of these attributes (their effects and interactions) is increasingly important in the present competitive and aggressive market. In fact, once the important attributes are determined, a company can adapt its product development strategy earlier than competitors, or more simply update their advertisement tactics. Although there has been a considerable improvement of models for predicting consumer behavior, the definition of practical methods able to efficiently translate theory into tools for preference capturing is still needed. Such consideration is due to the intrinsic complexity of the decision making science. This work aims at presenting a new practical method for capturing consumer attribute preferences indirectly, by using the choice time in a ranking task. It allows the analyst to indirectly obtain a respondent’s relative importance weights for several tested attributes by a simple, fast and economical procedure. Moreover, it allows overcoming most of the problems with context, survey and cognitive variables, i.e. variables influencing cognitive mechanisms driving consumers in their decision processes. A validation of the proposed method and its statistical consistency is illustrated through the results of a real experiment concerning the attributes of a cellular phone.
  • Revision of the Guide to the Expression of Uncertainty in Measurement

    Authors: Alistair Forbes
    Affiliation: National Physical Laboratory
    Primary area of focus / application:
    Keywords: Metrology, Reliability, Measurement Uncertainty
    Submitted at 25-Jul-2008 11:26 by Alistair Forbes
    Accepted
    23-Sep-2008 12:40 Revision of the Guide to the Expression of Uncertainty in Measurement
    Since its publication by ISO in 1995, the Guide to the Expression of Uncertainty in Measurement (GUM) has been successful in fostering a more coherent and probabilistic basis for reporting measurement uncertainty and it has been adopted by National Metrology Institutes and measurement laboratories throughout the world. The GUM does have its critics, some pointing to inconsistencies, others concerned with its perceived narrow scope or presentational issues. In order to broaden the scope, supporting documents are being prepared on various aspects, including multivariate measurands, conformance to specification and the use of Monte Carlo methods in propagating probability distributions, the latter document recently published as GUM Supplement 1.

    However, the GUM document itself is undergoing revision and it remains to be seen how far reaching the revision will be. The GUM, as it stands, can be seen as giving guidance on how to characterise the accuracy of a measurement system. Given a measurand (such as the length of a gauge block) it aims to describe the spread of results that could be expected for such a measurement system, due to various influence factors (such as temperature). Measurement system characterisation can be regarded as a type of forward uncertainty evaluation, from influence factors to measurement result. Users of measurement systems are interested in another question: given that the measurement system has produced the measured data, what can be said of the likely value of the measurand. For measurement systems with linear responses, inferences about the measurand can be made directly from the measurement system characterisation. For systems with nonlinear characteristics, the passage from measurement system characterisation to inferences about the measurand requires a type of inverse uncertainty evaluation, usually achieved through the application of Bayes’ theorem. The extent of the revision of the GUM will depend on to what extent the revised version encompasses both forward and inverse uncertainty evaluation. In this paper, we illustrate some of the issues in forward and inverse uncertainty evaluation for alignment errors in length metrology.