ENBIS-11 in Coimbra

4 – 8 September 2011 Abstract submission: 1 January – 25 June 2011

My abstracts

 

The following abstracts have been accepted for this event:

  • Identification of Adequate Reliability Verification Targets

    Authors: Nikolaus Haselgruber Franz Langmayr Christopher Gray
    Primary area of focus / application: Reliability
    Keywords: Reliability , Reliability Target , Testing , Product Development
    Submitted at 29-Apr-2011 13:27 by Nikolaus Haselgruber
    Accepted (view paper)
    5-Sep-2011 10:30 Identification of Adequate Reliability Verification Targets
    The reliability development of a complex technical product typically starts with analysis and simulation activities. Reliability risks which cannot be covered adequately with those virtual methods have to be addressed by physical experiments. The test units are (i) single components, (ii) subsystems or (iii) instances of the complete product, depending on the development level. To identify an appropriate experimental program for each level, corresponding verification targets have to be determined.
    The classical way is to decompose a given reliability target defined for the complete product according to its hierarchic structure. This approach has several shortcomings, e.g. that in particular the test effort required to demonstrate the targets derived for components and subsystems is usually too high and does not fit to given development budgets.
    The presentation shows an approach for the allocation of verification targets which considers a priori information from technical experts’ experience and, if available, field reliability observations of the forerunner product.
  • Improving Shewhart-type Control Charts for Monitoring Uni and Multivariate Process Variability

    Authors: Emanuel Pimentel Barbosa , Mario Antonio Gneri and Ariane Meneguetti
    Affiliation: Universidade Estadual de Campinas - UNICAMP
    Primary area of focus / application: Process
    Keywords: statistical process control , control charts , monitoring variability , Tukey studentized range , Cornish-Fisher correction , Meijer G-function
    Submitted at 29-Apr-2011 22:44 by Emanuel Pimentel Barbosa
    Accepted (view paper)
    6-Sep-2011 10:45 Improving Shewhart-type Control Charts for Monitoring Uni and Multivariate Process Variability
    The more frequently used statistics for monitoring process variability are the sample range R for the univariate case and the generalized variance |S| in the multivariate case, both of them usually implemented as a 3-sigma Shewart-type control chart based on normal approximation of the sample statistics.
    In this paper we show that this sort of “ad-hoc” procedures inflates substantially the alpha risk (probability of false alarm) and therefore should be avoided. Improved versions of these control charts are presented in this paper based on exact or better approximations of the sample statistic distributions and their quantiles (chart limits).
    For the R statistic we explore its relation with the Tukey’s maximum studentized range statistic distribution which provides simpler alternative background than the traditional Tippett`s integral formulae as well as it provides an easily way of implementation.
    In this paper the sample |S| statistic is better approximated using Cornish-Fisher quantile correction as well as it`s exact distribution is provided using the Mejer G-function, available in some recent mathematical software.
    The advantages of the proposed improved charts are shown in terms of alpha risk and ARL(Average Run Length) , and a couple of examples with real data are provided.
  • Likelihood Ratio Control Charts for Multivariate Process Variability Monitoring and Testing of Covariance Matrices

    Authors: Emanuel Pimentel Barbosa , Mario Antonio Gneri and Ariane Meneguetti
    Affiliation: Universidade Estadual de Campinas - UNICAMP
    Primary area of focus / application: Process
    Keywords: statistical process control , multivariate control charts , monitoring variability , likelihood ratio statistic
    Submitted at 30-Apr-2011 00:43 by Emanuel Pimentel Barbosa
    Accepted (view paper)
    5-Sep-2011 17:05 Likelihood Ratio Control Charts for Multivariate Process Variability Monitoring and Testing of Covariance Matrices
    Two very important and used statistics for monitoring and testing multivariate process variability, based on the sample covariance S matrix, are the generalized variance |S| and the modified likelihood-ratio LR ( or -2 lnLR ) statistic, the object of the present paper. Although the LR statistic has good theoretical properties, there is the challenge of obtaining sufficiently good approximations for its distribution and quantiles, since its exact distribution is practically untractable.
    In this paper, five different methods for approximating the -2 lnLR statistic distribution (basic chi-square, Satterthwaite-Bartlett, Box-Korin, Sugiura and Improved Sugiura) are compared in terms of false alarm risk for each method through simulation of ten million samples (based on Wishart random matrices, using Matlab). Also from this simulation study, one table of upper quantiles (alpha equal 0,0027) is produced for dimensions p = 2, 3 and 4 and sample sizes N from 3 to 10, 15, 20 and 30, which makes possible the practical implementation of the -2lnLR control chart. In order to illustrate the proposed mehod, a numerical example with real data is provided from a three dimensional quality characteristics of a production process.
  • Measurement Uncertainty - A Process Improvement Tool

    Authors: Phil Lewis
    Affiliation: Coventry University - Faculty of Engineering and Computing
    Primary area of focus / application: Metrology & measurement systems analysis
    Keywords: Measurement-Uncertainty , Design , Supply-chain , Profit-margin
    Submitted at 30-Apr-2011 17:15 by Phillip Lewis
    Accepted
    5-Sep-2011 16:45 Measurement Uncertainty - A Process Improvement Tool
    The existence of measurement uncertainty is being highly regarded at the top end of industry and metrology as vital in the safety critical environment sectors such as radiotherapy, nuclear power and aerospace. The concept is creating extensive knowledge and understanding to the entire envelope of the relevant sectors.

    Within these business sectors the availability of statisticians and expert practitioners is freely available. Once the step is made into the various supply chains of any business sector the culture which has been created by the lean operations drive has promoted all forms of verification and measurement as “non value added”. Thus the concept that a NVA process also has an unknown has provided an excuse for it to be ignored.

    An investigation into the current accepted practice within the SME activity would unveil many instances of disagreements between entities of the supply chain where contrasting measurements have been obtained and the resolution over which one is correct erodes reduced profit margins frantically.

    The emerging trend of designing and managing processes “up stream” must be suitable for the introduction of “design for measurement uncertainty” which would introduce the process improvement concept into the measurement arena to statistically evaluate confidence levels within any organisation measurement systems and thus align the process capability of measurement systems and their alignment from interacting department right through to OEM’s partnerships with second tier suppliers and trade dealers.
  • Incomplete jigsaw puzzle model of problem solving process

    Authors: Jan M. Msyzewski
    Affiliation: Kozminski University, Warsaw
    Primary area of focus / application: Quality
    Keywords: knowledge management, , problem solving , data transfer , jigsaw puzzle solving
    Submitted at 1-May-2011 21:01 by Jan Myszewski
    Accepted
    5-Sep-2011 16:45 Incomplete jigsaw puzzle model of problem solving process
    The purpose of this paper is to discuss the problem solving process from the view point of information transfer.
    A process of problem solving is represented by the metaphor of solving jigsaw puzzle, with some pieces missing. In particular, certain number are missing permanently.
    Problem solving is a special case of process of the knowledge management in organization. Important part are processes of data transfer between participants. Model of incomplete jigsaw puzzle enables describing these transfers by use of some probabilistic distributions.
    The model enables discussion of variants of strategies of problem solving and their association with organizations culture and knowledge management.
  • A Comparison of Decision Procedures for Cpk when Data are Autocorrelated

    Authors: Peder Lundkvist Kerstin Vännman Murat Kulahci
    Primary area of focus / application: Quality
    Keywords: process industry , continuous process , autocorrelation , capability index Cpk
    Submitted at 5-May-2011 15:41 by Peder Lundkvist
    Accepted (view paper)
    5-Sep-2011 17:10 A Comparison of Decision Procedures for Cpk when Data are Autocorrelated
    Autocorrelated data in industry is becoming increasingly common due to, for example, on-line data collection systems with high-frequency sampling. Thus, the basic assumption of independent observations is not fulfilled. However, most decision procedures for process capability analysis assume independent data. A consequence of the autocorrelation is that the distribution of the estimated index is unknown, which affects the decision rule. A handful of methods have been suggested to deal with the problem of autocorrelated data in process capability analysis.
    The purpose of the presentation is to compare the performance of three different decision rules in process capability analysis using the process capability index when data are autocorrelated. This is done through a case study followed by a simulation study. In the case study the data is obtained for the weight percentage of the carbon content in pig iron from a blast furnace process, and can be described by an AR(1)-model. In the simulation study the significance level and power of each method are investigated.